Soybean Crush Spread Arbitrage:

Trading Strategies and Market Efficiency

Presented at the Academy of Finance

Chicago, Illinois
March 28, 2007

John B. Mitchell
328 Sloan Hall
Central Michigan University
Mt. Pleasant, MI 48859
989 774 3651
mitch1jb@cmich.edu

Electronic copy available at: http://ssrn.com/abstract=987507

Soybean Crush Spread Arbitrage:

Trading Strategies and Market Efficiency
John B. Mitchell1
Central Michigan University
Abstract
This paper revisits the soybean crush spread arbitrage work of Simon (JFM, 1999). Major
findings are that contrary to the results reported by Simon, the length of winning and losing
trades differ systematically. Winning trades are significantly shorter on average than losing
trades. This result leads to trading rules designed to prevent lengthy trades. Secondly, the work
by Simon employs symmetric entry and exit limits. That approach assumes that the market
overshoots equilibrium. This research studies a wide variety of entry and exit limits and the riskreturn relationship between the entry and exit limits.
Literature Review
Simon (1999) studies the soybean futures crush spread (soybean meal and oil versus soybeans)
over the period January, 1985 though February, 1995. He finds that the crush spread tends to
revert toward its 5-day moving average with a half-life of deviations of about 3 days. He
proposes trading rules to take advantage of this relationship. Simon studies entry and exit limits
of $100, 200 and 300 above and below the recent 5-day moving average. Trades are reversed
when the spread reaches a level equal to, but of opposite sign, to the initiating trigger. For
example, a trade to sell the spread is initiated when the spread is $100 larger than its 5-day
moving average and reversed when the spread is $100 less than its 5-day moving average. This
procedure counts on a positive spread (relative to the recent average) becoming a negative spread
(again relative to its trailing average).
Trades prove profitable after adjustment for transaction costs ($103.50 per round-trip trade) with
the average profit ranging from $27 to $355 for long trades as the limits are increased from $100
to $300 and similarly from $33 to $185 for short trades. Approximately two thirds of all trades
prove profitable. Trading rules based on a regression-derived fair market value improve both the
percentage of profitable trades and average profit while reducing the number of trading
opportunities to as few as 3 over a 10-year period. The fair-value consideration also adds to the
complexity of determining when to initiate a trade. Simon also reports that the average length of
winning and losing trades does not differ significantly.
Rechner and Poitras (1993) examine intraday trading opportunities in the soy complex during the
period of 1978-1991. They find that nave spread reversal trades are profitable when a filter is
applied. The profit per trade increases, after transaction costs of 1.5 cents per bushel, from a loss
of .35 cents per bushel with no filter to +.35, 1.02, and 1.74 respectively at filters of 1, 2, and 3
1

Dr. John B. Mitchell is Professor of Finance at Central Michigan University. This project was supported by a
sabbatical leave. He thanks G. Thomas Mitchell of Aurora Consulting for programming assistance. Presented at the
Academy of Finance, Chicago, Illinois, March 28, 2007. All rights reserved.

2
Electronic copy available at: http://ssrn.com/abstract=987507

cents per bushel. The percentage of profitable trades increases from 39.6% to 69.0 % over the
same range. The coefficient of variation, derived from their results, is 8.49, 2.83, and 2.30 for the
same three filters. They also report significant positive skewness (near 1) and kurtosis (6 to 7) of
the distribution of trade results.
Johnson, Zulauf, Irwin, and Gerlow (1991) explore various fixed-length trades for the soybean
crush spread over the period 1966-1988. Deviations of an implied net crushing margin from its
60-day moving average potentially trigger trades on the 15th of each month. Trade lengths of 1.5,
3.5, 5.5, 7.5, and 9.5 months are studied. Transaction costs of $207 are deducted from profits for
trade lengths of 5.5 months or less and $236 for longer trades. Trades are reversed on the 1st
trading day of the month.
Johnson, et al find that longer trade lengths and larger filters (deviations from recent average)
result in larger profits and a higher percentage of profitable trades. While 1.5 month trades result
in an average loss of $156, profit increases to $3, $241, $235, and $418 as the trade length
increases. Filters of 20 cents result in near 100% profitable trades, but very few, 2 to 24, trades
over the study period. Filters of 10-20 cents result in 38% to 60% profitable short (sell the
spread) trades and 57% to 65% profitable long trades, but still only 10 to 39 trades over the study
period. Smaller filters are even less successful. Traders are left with the choice of very few, but
profitable, trades at high filters, or a low probability of success and high coefficient of variation
at lower filters.
Numerous authors provide results from other financial and commodity futures markets. Barrett
and Kolb (1995) find little evidence of regularities that would lead to profit-making opportunities
in a wide variety of intra and inter-market commodity spreads. Girma and Paulson (1999) test
various magnitude deviations, measured in standard deviations, from the moving average for the
petroleum crack spread and utilize movement beyond the 5-day moving average for an exit
trigger. Castelino and Vora (1984) find that volatility of the spread increases with spread length.
Poitras (1998) discusses whether the existence of profits necessarily means market inefficiency
due to variations in transaction costs and liquidity constraints, especially in longer maturity
contracts. He tests fixed length trades in the TED spread during the period 1983-1991 and finds a
small number of profitable (before transaction costs) trades (generally 1 to 4 per year),
improvement in profit and percentage of profitable trades if a filter is applied, and that the day of
the month that trades occur affects profitability. Ma and Soenen (1988) find positive riskadjusted profits in the gold silver spread over the period 1976-1986 using next-day prices to
ensure market access. They employ 28 and 30-day moving averages and standard deviation
based filters to generate approximately 20 trades per year with more than 50% profitable. A 30day moving average and .5 standard deviation of the spread filter result in an average (unweighted) coefficient of variation of 3.04 and 61% profitable trades. Wahab, Cohn, and Lashgari
(1994) also study the gold-silver spread and find negative returns after transaction costs. Some
infrequent moving average signals result in highly risky profits. Carter (1989) studies spreads
between liquid versus illiquid futures contracts (corn versus barley, soybeans versus rapeseed,
and Chicago versus Toronto T-bonds) and finds mixed results with a limited number of
transactions. Only the T-bond spread proves profitable. Elfakhani and Wionzak (1997) update
the Carter study and find limited statistically insignificant arbitrage opportunities.

More recently, Dunis, Laws, and Evans (2006) update the work of Butterworth and Holmes
(2002). Dunis et al find that a moving average model outperforms a fair value model for
petroleum inter-market spreads from 1995-2004. They find annualized returns as high as 26.15%
using a correlation filter and including transaction costs. Similarly, Haigh and Holt (2002) find
that a multivariate GARCH model increases the effectiveness of crack spread hedging.
The interest in the risk reduction and arbitrage opportunities in spread trading is exemplified by
the recent introduction of spread futures on the Chicago Board of Trade. This development is
chronicled by Cuny (2006). As this paper is being written the CBOT has introduced Soybean
Crush Options, further emphasizing the need existing in the market for understanding the
behavior of the soybean crush spread.

Motivation for Study

This study differs from earlier research in the following ways:
1)

This study reverses trades at, or short of, equilibrium rather than beyond equilibrium
spread value. Simon finds mean reversion. However, mean reversion does not necessarily
imply that the spread will continue beyond equilibrium in an oscillating manner. By
choosing closing limits short of equilibrium, this study attempts to increase the
percentage of profitable trades and reduce the standard deviation of profit; albeit at the
possible expense of reduced profit.

2)

Truncating trades is explored as a way to avoid adverse trade results. This truncation is
based on the observation, discussed in the results section, that losing trades are, on
average, significantly longer than winning trades.

3)

A much longer time period is studied. This study covers almost 20 years while Simon
studies 10 years. Some trading strategies result in relatively few trades per year and thus a
larger sample of simulated trades is produced. This study also serves as an out-of-sample
replication of Simon.

4)

A larger number of measures of performance are studied in an attempt to better

understand the nature of any arbitrage opportunities.

5)

Trading strategies are based on simple indicators derived only from futures market data
readily available to traders. Simon also employs a fair market value concept in an attempt
to improve arbitrage performance. This study employs only simple comparisons readily
available to traders on the floor.

Data
Soybean, Soybean Meal, and Soybean Oil futures contract prices collected from Commodity
Systems Inc. are studied for the period from June 4, 1984 through April 7, 2006. The closing

price for each trading day is employed to calculate the crush spread with the contracts rolled over
on the first day of the month before expiration. All three contracts have January, March, May,
July, August, and September expirations. Soybeans also have a November contract while oil and
meal have October and December contracts. The same contract expirations are employed for
each commodity with the exception of trading November soybeans against December oil and
meal. For example, on the first trading date in April, the May contracts are rolled over to the
June contract. On the first trading date in August, the September contract in soybeans is rolled
over to the November contract and the oil and meal contracts are rolled over to the December
contracts. This methodology is the same as that employed by Simon.

Crush Spread
The crush spread is calculated using one futures contract for each commodity2. Because of the
difference in contract specifications between the three commodities the crush spread is calculated
as:
Crush Spread = (Soybean Meal in $ per ton x 100) + (Soybean Oil in $ per 100 lbs. x 600)
(Soybeans in cents per bu. x 50)
(1)

Trading Rules
Limits (deviations from the most recent 5-day (see Simon) moving average of up to +/- $400) are
set for establishing short and long crush spread transactions as follows:
If the crush spread exceeds the upper limit, sell the spread i.e. sell oil and meal and buy
soybeans.
If the crush spread falls below the lower limit, buy the spread i.e. buy oil and meal and sell
soybeans.
Reverse (close) the transaction if the spread moves beyond closing limits. Closing limits between
zero and $200 are tested. This rule differs from Simon. Simon tests movement beyond an
equilibrium level. Mean reversion does not necessarily imply that the spread will do anything
more than revert towards the mean. There is no reason to expect that the spread will continue in
an oscillating manner beyond the equilibrium level. Once exceeding equilibrium level there
should be immediate pressure to move back towards equilibrium. Therefore, it is unreasonable to
expect, and indeed count upon, continued movement away from equilibrium. Closing limits short
of equilibrium are tested to examine if stopping short improves the percentage of profitable
trades and other trade characteristics.
Closing transactions on rollover dates are triggered based on the previous contracts while
opening transactions on those dates are triggered based on the new contracts. Therefore, the 52

A 1:1:1 spread does not exactly match contract sizes. Secondly, the relationships change over time due to soybean
varieties and growing conditions. See Simon (1999), Johnson et. al. (1991), and the CBOT website for elaboration.

day moving average is always calculated based on the previous 5 days for the same contract
expiration month being employed to trigger a transaction.
Only one trade is allowed to exist at any given point in time. New positions are not created until
existing trades have been reversed. The use of symmetric limits precludes the establishment of
long positions while short positions, or vice versa, are outstanding.

Results
Results are presented without transaction costs for both short and long trades for each
combination of trade limits. For ease of reporting the trade limits are symmetric, i.e. long and
short trades use the same limit with opposite sign, and limits are reported as opening/closing.
Results consist of the number of trades, average profit, standard deviation of profit, coefficient of
variation, percentage of profitable trades, maximum profit and loss, average length of trade,
average length of winning and losing trades, maximum length of winning and minimum length
of losing trades, and maximum trade length.
Number of Trades
The number of trades (see Table 1) over the 20-year study varies from a low of 70 in the case of
400/0 thru 50 (an opening trade limit of 400 and a closing limit of anywhere between 0 and 50)
to a high of 1,257 for 40/0. The effects of increasing the closing limit are relatively modest. For
example, at an opening limit of 400, raising the closing limit from 0 to 200 only increases the
number of trades by 6% (from 70 to 74) while at an opening limit of 220, raising the closing
limit to 200 increases the number of trades by 20% (from 285 to 341). Closing limits of half the
size of opening limits result in an approximate 13% increase in trades (relative to a return to the
5-day moving average) at opening limits below 220 and lesser amounts at higher opening limits.
This increase in the number of trades as the closing limit is increased is a result of completing
transactions more rapidly and therefore clearing the way for additional trades.
Table 1
Number of Trades
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
1257
1099
954
805
682
578
478
403
340
285
238
209
183
148
131
117
99
82
70

10
1311
1138
981
827
700
592
487
410
347
291
242
211
185
149
132
118
99
82
70

20
1368
1181
1009
846
711
598
490
412
349
292
243
211
185
149
132
118
99
82
70

30
1438
1234
1047
871
729
612
498
417
351
294
243
211
185
149
132
118
99
82
70

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

1285
1080
892
739
619
501
419
352
295
243
211
185
149
132
118
99
82
70

1326
1109
911
751
627
508
424
355
297
245
212
186
150
133
118
99
82
70

1145
935
769
642
518
431
362
301
248
214
187
151
134
119
100
83
71

1177
964
786
650
525
435
365
304
249
215
188
152
135
119
100
83
71

994
811
669
538
444
370
307
252
217
190
154
137
121
102
83
71

1018
831
686
547
451
376
313
256
220
192
154
137
121
102
83
71

844
694
554
455
379
316
258
221
192
154
137
121
102
83
71

867
709
565
462
385
320
260
223
193
154
137
121
102
83
71

723
576
468
389
322
260
223
193
154
137
121
102
83
71

733
582
472
391
323
261
224
194
155
138
122
102
83
71

592
482
395
326
263
226
195
156
139
123
103
84
72

601
487
400
330
265
228
196
157
140
124
104
84
72

492
402
331
266
229
196
157
140
124
104
84
72

494
404
332
267
229
196
157
140
124
104
84
72

411
336
271
232
199
160
142
126
105
85
73

412
337
271
232
199
160
142
126
105
85
73

341
275
235
201
162
143
127
106
86
74

The number of trades is much more sensitive to the opening limit. Each increase of 20 in the
opening limit, from 40 to 400, results on average in a 5.25% reduction in the number of trades.

However, the reduction begins much more rapidly at lower opening limits. When considering the
number of trades it is important to note that the reported results are for a 20-year period. Even
the maximum number of 1,257 amounts to no more than one trade every 6 days; and at the
highest limits only one trade every 104 days. Unless trades prove very profitable, only an
automated trading system could prove economically worthwhile.
Average Profit
Average profit (see Table 2) ranges from a low of $63.54 at limits of 40/0 to $225.67 at 400/80.
The maximum profit for each opening limit generally occurs at, or near, a closing limit near zero.
Closing limits half as large as opening limits, result on average in a 15% decrease in average
profit with a decrease of 21% at 340/170. Increasing the opening limit increases average profit
rapidly between 140 and 320 with more modest increases at each end of the range.
Table 2
Average Profit ($) No transaction cost
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
64
67
72
77
82
88
100
108
115
129
145
162
180
191
209
204
215
223
223

10
62
66
71
77
82
89
100
109
116
130
144
159
176
187
205
200
210
218
216

20
59
63
67
74
79
84
95
104
110
124
137
151
168
179
198
193
204
212
209

30
56
60
65
72
78
83
93
102
108
123
136
149
168
177
195
190
204
212
209

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

57
62
68
75
79
91
98
104
118
132
145
164
173
192
187
203
210
208

54
59
65
72
77
89
96
103
118
132
145
164
175
193
190
207
215
214

57
65
71
76
88
96
102
115
127
141
159
169
187
185
204
222
223

55
63
69
74
87
95
101
114
126
141
160
170
188
184
203
222
225

62
68
73
86
93
99
113
125
140
160
170
189
186
204
222
226

61
68
74
85
93
100
114
126
141
161
167
185
181
199
216
218

66
72
83
91
98
112
124
140
159
166
185
181
199
215
217

64
70
81
88
95
108
119
136
154
159
177
173
192
204
207

70
81
88
95
108
118
135
153
156
175
171
190
201
203

69
79
86
94
106
117
134
153
157
176
173
192
202
202

79
87
94
107
117
134
152
157
175
173
192
203
202

79
87
94
108
118
134
152
157
176
173
193
204
204

81
88
100
110
128
145
151
170
166
185
199
198

80
86
97
107
125
142
147
166
162
180
192
190

89
100
109
129
145
152
171
169
188
202
202

87
97
106
126
142
149
166
163
181
193
192

95
104
122
138
146
164
159
179
190
188

Standard Deviation of Profit

The standard deviation of profit (see Table 3) ranges from a low of $209.20 at 40/30 to a high of
$595.38 at 400/0. The standard deviation increases as the opening limit is increased and as the
closing limit is decreased. The standard deviation is less sensitive to changes in the closing limit,
at least at low levels of the closing limit, when the opening limit is large. This suggests that
Table 3
Standard Deviation of Profit ($)
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
228
239
250
265
280
299
312
327
342
360
384
399
419
452
474
499
528
556
595

10
221
232
245
259
274
292
306
321
334
352
377
393
413
447
469
494
526
554
593

20
215
227
240
255
270
289
304
320
333
352
375
392
412
446
468
493
526
554
593

30
209
221
234
249
265
284
300
316
330
348
374
390
410
445
468
493
526
554
593

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

216
230
246
263
282
300
316
330
348
374
391
411
446
468
493
526
554
592

211
225
241
259
279
296
312
329
347
373
389
409
443
465
491
523
551
590

213
229
245
263
282
297
312
332
356
369
392
424
444
468
500
513
548

210
225
242
261
280
296
311
330
357
370
392
423
443
469
500
513
547

220
236
255
274
291
306
326
351
365
386
416
435
459
489
509
543

212
228
246
267
283
297
316
341
355
376
413
431
456
485
504
538

225
243
265
281
296
314
339
353
375
412
430
455
485
504
538

223
242
263
280
295
313
339
353
377
415
433
458
488
509
543

238
260
277
293
311
338
352
376
415
433
457
488
508
542

236
257
275
291
311
337
351
375
413
430
455
487
508
542

257
274
292
313
342
356
381
420
439
463
496
518
553

253
272
289
309
339
352
378
416
435
459
491
518
553

268
286
306
336
352
378
416
436
460
492
520
553

272
291
313
343
359
386
425
445
470
503
529
563

253
271
296
307
329
360
376
395
421
430
453

249
267
292
302
323
353
370
389
414
422
445

267
290
301
323
351
369
388
413
420
442

closing limits near zero may result in more advantageous risk-return relationships. This is
explored later by examining the coefficient of variation.
Coefficient Variation
The coefficient of variation (see Table 4) ranges from a low of 2.13 at 380/180 to a high of 3.91
at 60/50. The coefficient of variation decreases in an irregular manner as the opening limit
increases and then rises at high levels of the opening limit. The lowest value for the coefficient of
variation occurs at opening limit levels of 280 to 380. The coefficient of variation is more
predictably u-shaped as the closing limit varies. At opening limits roughly below 200 the
minimum coefficient of variation occurs at a closing limit of 10. At higher opening limits the
minimum shifts to a closing limit of 180. The coefficient is fairly steady over a wide range of
limits suggesting a consistent risk-return relationship.
Table 4
Coefficient of Variation
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
3.58
3.54
3.50
3.43
3.43
3.40
3.12
3.02
2.97
2.78
2.64
2.46
2.33
2.36
2.27
2.45
2.46
2.49
2.67

10
3.56
3.54
3.47
3.38
3.32
3.29
3.06
2.95
2.87
2.70
2.61
2.47
2.34
2.39
2.29
2.47
2.51
2.54
2.74

20
3.63
3.63
3.56
3.46
3.42
3.44
3.19
3.09
3.02
2.83
2.73
2.60
2.45
2.50
2.37
2.55
2.57
2.62
2.84

30
3.71
3.69
3.62
3.48
3.41
3.42
3.22
3.11
3.07
2.84
2.75
2.61
2.44
2.51
2.40
2.59
2.57
2.62
2.84

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

3.80
3.74
3.60
3.52
3.57
3.30
3.21
3.18
2.94
2.83
2.70
2.51
2.57
2.44
2.64
2.59
2.64
2.84

3.91
3.83
3.69
3.61
3.64
3.32
3.24
3.19
2.94
2.83
2.69
2.49
2.54
2.41
2.59
2.53
2.57
2.75

3.74
3.54
3.47
3.48
3.20
3.09
3.05
2.88
2.81
2.62
2.47
2.51
2.37
2.53
2.45
2.31
2.45

3.82
3.57
3.51
3.54
3.23
3.13
3.09
2.90
2.82
2.62
2.45
2.49
2.36
2.55
2.46
2.30
2.43

3.56
3.46
3.47
3.20
3.12
3.09
2.88
2.80
2.61
2.42
2.45
2.30
2.48
2.39
2.29
2.41

3.48
3.36
3.33
3.14
3.06
2.97
2.77
2.71
2.51
2.33
2.48
2.33
2.52
2.44
2.33
2.46

3.41
3.37
3.18
3.07
3.01
2.80
2.75
2.52
2.35
2.48
2.32
2.51
2.44
2.35
2.48

3.49
3.45
3.25
3.17
3.11
2.91
2.85
2.60
2.45
2.62
2.44
2.64
2.54
2.49
2.63

3.40
3.21
3.14
3.08
2.89
2.86
2.61
2.46
2.65
2.47
2.67
2.57
2.53
2.67

3.44
3.26
3.19
3.11
2.93
2.89
2.61
2.46
2.64
2.45
2.63
2.54
2.51
2.69

3.25
3.16
3.11
2.92
2.92
2.66
2.50
2.68
2.50
2.68
2.58
2.56
2.73

3.19
3.12
3.06
2.87
2.88
2.62
2.48
2.65
2.47
2.65
2.55
2.54
2.71

3.29
3.25
3.07
3.07
2.74
2.61
2.76
2.57
2.77
2.66
2.62
2.80

3.42
3.38
3.21
3.22
2.87
2.72
2.89
2.69
2.91
2.80
2.75
2.96

2.83
2.72
2.71
2.39
2.26
2.36
2.19
2.34
2.24
2.13
2.25

2.87
2.75
2.74
2.40
2.28
2.38
2.22
2.38
2.29
2.18
2.32

2.82
2.79
2.48
2.34
2.40
2.26
2.43
2.31
2.21
2.35

Percentage of Profitable Trades

The percentage of profitable trades (see Table 5) ranges from 72.11% at 200/50 to 79.76% at
380/140 and 150. The percentage is fairly steady but tends to be slightly lower around opening
limits of 160-200.
Table 5
Percentage of Profitable Trades
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
76%
76%
75%
75%
74%
74%
74%
75%
73%
73%
75%
77%
78%
77%
79%
76%
78%
77%
77%

10
76%
76%
75%
75%
75%
75%
75%
75%
73%
74%
75%
76%
78%
77%
79%
76%
78%
77%
77%

20
77%
76%
75%
75%
75%
75%
74%
74%
72%
73%
74%
76%
78%
77%
79%
76%
78%
77%
77%

30
76%
76%
75%
76%
75%
75%
74%
74%
72%
73%
74%
76%
78%
77%
78%
75%
78%
77%
77%

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

77%
76%
77%
76%
75%
74%
73%
72%
73%
74%
75%
77%
75%
77%
74%
77%
77%
77%

76%
76%
76%
75%
75%
74%
73%
72%
73%
74%
75%
77%
75%
77%
75%
78%
78%
79%

76%
76%
75%
75%
74%
74%
73%
73%
74%
75%
77%
74%
75%
73%
77%
77%
77%

76%
77%
76%
76%
75%
74%
73%
73%
73%
75%
77%
74%
74%
72%
77%
77%
77%

77%
77%
76%
75%
75%
74%
74%
74%
75%
77%
75%
75%
74%
78%
78%
79%

77%
78%
77%
75%
75%
75%
76%
75%
77%
79%
75%
75%
74%
78%
78%
79%

77%
77%
75%
75%
75%
76%
76%
77%
79%
75%
76%
74%
78%
78%
79%

76%
76%
75%
75%
74%
75%
74%
75%
77%
73%
76%
74%
78%
78%
77%

76%
76%
75%
74%
75%
74%
76%
77%
73%
76%
74%
78%
78%
77%

76%
76%
75%
74%
76%
74%
76%
78%
75%
78%
76%
79%
80%
77%

76%
76%
75%
76%
75%
77%
77%
74%
76%
76%
80%
80%
78%

76%
76%
76%
77%
75%
77%
78%
74%
76%
75%
79%
80%
78%

76%
75%
77%
76%
76%
77%
73%
76%
75%
79%
79%
76%

75%
75%
76%
75%
76%
77%
73%
76%
75%
79%
79%
76%

76%
77%
76%
78%
78%
74%
76%
75%
78%
78%
75%

75%
76%
75%
78%
78%
74%
76%
75%
78%
78%
75%

76%
76%
78%
79%
76%
78%
76%
78%
78%
76%

Maximum Profit
The maximum profit is $1,314 at all limits.
Maximum Loss
The maximum loss is -$2,986 at all closing limits of 130 or lower, -$3,146 at all closing limits of
140 through 170, and -$1,411 at closing limits of 180-200 combined with opening limits of 360
or less. Combinations of closing limits of 180-200 and opening limits of 380-400 result in a
maximum loss of -$1,350. With the exception of opening limits above 360, all opening limits
result in the same maximum loss at each closing limit level, i.e. the closing limit controls the
maximum loss within the range of limits studied.
Average Trade Length
The average trade length (see Table 6) ranges from 1.79 days at 220/200 to 4.34 days at 200/0.
At low closing limit levels the average trade length generally increases slightly with the opening
limit to an opening limit of 200, and then generally declines slightly. However, at closing limit
levels of 60 or higher the average trade length continues to increase as the opening limit
increases. Overall, average trade length exhibits relative stability.
Table 6
Average Trade Length (Days)
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
3.72
3.88
4.01
4.12
4.24
4.29
4.31
4.32
4.34
4.30
4.17
4.15
4.08
4.07
4.02
4.15
4.04
3.90
3.94

10
3.45
3.61
3.77
3.89
3.97
4.05
4.08
4.11
4.11
4.09
3.98
3.98
3.91
3.93
3.88
3.98
3.95
3.82
3.84

20
3.18
3.34
3.49
3.62
3.71
3.82
3.89
3.95
3.95
3.93
3.84
3.81
3.78
3.85
3.81
3.92
3.90
3.76
3.77

30
2.90
3.05
3.21
3.34
3.44
3.55
3.64
3.70
3.73
3.72
3.70
3.68
3.64
3.77
3.76
3.86
3.87
3.72
3.73

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

2.80
2.96
3.12
3.25
3.37
3.48
3.56
3.59
3.57
3.57
3.56
3.52
3.66
3.66
3.75
3.76
3.68
3.70

2.60
2.76
2.92
3.06
3.18
3.29
3.37
3.44
3.45
3.44
3.47
3.44
3.55
3.56
3.63
3.63
3.52
3.57

2.57
2.73
2.87
2.98
3.08
3.14
3.20
3.28
3.27
3.29
3.26
3.37
3.37
3.45
3.48
3.36
3.48

2.40
2.55
2.70
2.84
2.93
3.01
3.08
3.16
3.17
3.20
3.17
3.28
3.26
3.39
3.42
3.30
3.42

2.36
2.50
2.63
2.74
2.85
2.92
3.00
3.02
3.04
3.03
3.11
3.09
3.21
3.23
3.13
3.27

2.22
2.35
2.47
2.60
2.70
2.76
2.83
2.87
2.91
2.89
3.03
3.00
3.10
3.15
3.07
3.20

2.23
2.35
2.48
2.58
2.66
2.73
2.77
2.81
2.81
2.94
2.92
3.01
3.04
3.02
3.14

2.06
2.18
2.31
2.42
2.50
2.55
2.61
2.65
2.65
2.79
2.76
2.83
2.86
2.89
3.00

2.07
2.19
2.31
2.38
2.44
2.53
2.57
2.56
2.71
2.70
2.81
2.83
2.86
2.97

1.97
2.10
2.23
2.30
2.37
2.45
2.49
2.49
2.64
2.62
2.72
2.75
2.81
2.96

1.98
2.10
2.20
2.26
2.37
2.41
2.42
2.57
2.55
2.65
2.70
2.74
2.88

1.91
2.03
2.12
2.17
2.29
2.32
2.34
2.48
2.49
2.58
2.62
2.71
2.85

1.92
2.00
2.05
2.16
2.21
2.25
2.40
2.41
2.50
2.53
2.65
2.79

1.86
1.94
2.00
2.10
2.15
2.19
2.35
2.36
2.45
2.47
2.60
2.72

1.86
1.92
2.01
2.06
2.08
2.24
2.27
2.37
2.40
2.51
2.62

1.80
1.86
1.95
2.01
2.02
2.19
2.24
2.33
2.35
2.45
2.55

1.79
1.87
1.93
1.94
2.11
2.15
2.26
2.29
2.37
2.46

Average Length of Profitable Trades

The average length of profitable trades (see Table 7) ranges from 1.30 days at 220/200 to 3.09
days at 220/0. The average length of winning trades behaves much the same as the average
length of all trades; as might be expected given the high percentage of profitable trades.
However, the average length of profitable trades is lower at all levels. This is the first indication
that winning and losing trades differ systematically in length.

Table 7
Average Length of Profitable Trades (Days)
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
2.44
2.58
2.68
2.77
2.84
2.87
2.95
3.03
3.05
3.09
3.03
3.01
2.96
2.96
2.93
2.92
2.78
2.63
2.59

10
2.26
2.40
2.53
2.65
2.73
2.77
2.82
2.91
2.92
2.97
2.95
2.91
2.88
2.89
2.87
2.84
2.71
2.59
2.54

20
2.09
2.21
2.34
2.45
2.52
2.60
2.65
2.74
2.74
2.77
2.77
2.73
2.74
2.79
2.80
2.78
2.68
2.54
2.48

30
1.91
2.02
2.14
2.25
2.33
2.42
2.46
2.53
2.56
2.64
2.68
2.64
2.65
2.73
2.73
2.70
2.65
2.51
2.44

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

1.87
1.98
2.11
2.21
2.27
2.35
2.41
2.45
2.52
2.56
2.51
2.53
2.60
2.64
2.62
2.59
2.49
2.43

1.74
1.85
1.97
2.08
2.15
2.23
2.30
2.35
2.45
2.49
2.48
2.52
2.58
2.63
2.60
2.57
2.47
2.44

1.73
1.86
1.97
2.02
2.10
2.18
2.25
2.34
2.37
2.37
2.36
2.39
2.40
2.40
2.45
2.36
2.44

1.63
1.74
1.85
1.92
1.99
2.08
2.14
2.25
2.26
2.28
2.26
2.29
2.28
2.33
2.39
2.30
2.38

1.62
1.72
1.79
1.86
1.97
2.03
2.14
2.16
2.15
2.16
2.17
2.18
2.22
2.29
2.20
2.32

1.56
1.64
1.72
1.80
1.91
1.97
2.06
2.07
2.10
2.08
2.08
2.08
2.10
2.21
2.14
2.25

1.55
1.63
1.71
1.81
1.88
1.97
1.98
2.02
2.03
2.03
2.04
2.06
2.08
2.08
2.18

1.44
1.50
1.57
1.66
1.73
1.79
1.84
1.88
1.87
1.88
1.93
1.94
1.98
2.00
2.07

1.42
1.49
1.58
1.63
1.68
1.75
1.80
1.79
1.80
1.87
1.92
1.95
1.97
2.05

1.37
1.43
1.52
1.57
1.63
1.70
1.75
1.76
1.78
1.84
1.90
1.93
1.97
2.05

1.38
1.46
1.51
1.57
1.64
1.69
1.70
1.71
1.76
1.82
1.88
1.91
1.98

1.36
1.42
1.47
1.53
1.59
1.64
1.64
1.66
1.71
1.75
1.80
1.88
1.95

1.36
1.39
1.43
1.49
1.52
1.53
1.57
1.62
1.68
1.73
1.82
1.89

1.34
1.36
1.40
1.45
1.49
1.49
1.53
1.58
1.63
1.68
1.76
1.82

1.35
1.38
1.43
1.46
1.46
1.51
1.56
1.62
1.67
1.74
1.80

1.31
1.35
1.38
1.42
1.41
1.48
1.52
1.58
1.62
1.68
1.73

1.30
1.31
1.35
1.35
1.42
1.45
1.51
1.55
1.60
1.63

Average Length of Losing Trades

The average length of losing trades (see Table 8) ranges from 3.24 days at 200/190 to 8.50 days
at 400/0. The average length of losing trades is always at least 2.42 times as large as the average
length of winning trades and on average (equally weighted for all combinations of limits) is 2.84
times as large. The average length of winning trades at limits of 380/140, for example, is
significantly less than the average length of losing trades at .01%. The opportunity to distinguish
between winning and losing trades on the basis of length, however, depends on how the two
distributions overlap.
Table 8
Average Length of Losing Trades (Days)
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
7.82
7.99
8.02
8.13
8.23
8.31
8.30
8.10
7.84
7.64
7.55
7.86
7.95
7.79
8.04
8.04
8.45
8.11
8.50

10
7.32
7.51
7.58
7.66
7.72
7.83
7.84
7.72
7.41
7.28
7.13
7.42
7.54
7.47
7.64
7.64
8.27
7.89
8.25

20
6.80
6.95
7.05
7.16
7.22
7.44
7.48
7.45
7.14
7.06
6.94
7.28
7.46
7.41
7.57
7.57
8.18
7.79
8.13

30
6.18
6.45
6.52
6.74
6.81
6.99
7.05
7.05
6.80
6.66
6.68
7.02
7.25
7.29
7.41
7.41
8.14
7.74
8.06

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

5.91
6.11
6.43
6.49
6.63
6.68
6.71
6.54
6.33
6.38
6.72
6.90
6.86
6.97
6.90
7.61
7.63
8.00

5.38
5.62
5.95
6.04
6.24
6.23
6.23
6.26
6.12
6.14
6.43
6.57
6.51
6.61
6.63
7.32
7.28
7.73

5.22
5.53
5.61
5.88
5.85
5.81
5.78
5.86
5.84
6.06
6.28
6.18
6.21
6.31
6.91
6.74
7.06

4.87
5.25
5.43
5.70
5.71
5.67
5.66
5.67
5.70
5.93
6.16
6.05
6.06
6.15
6.87
6.68
7.00

4.90
5.16
5.33
5.38
5.45
5.38
5.48
5.51
5.72
5.98
5.87
5.85
5.94
6.64
6.50
6.80

4.45
4.82
4.97
5.04
5.14
5.13
5.28
5.32
5.59
5.85
5.82
5.79
5.88
6.55
6.44
6.73

4.49
4.75
4.83
4.94
5.00
5.16
5.21
5.50
5.75
5.71
5.70
5.77
6.55
6.44
6.73

4.08
4.39
4.57
4.66
4.66
4.79
4.79
4.98
5.20
5.27
5.36
5.42
6.09
6.11
6.19

4.16
4.40
4.52
4.54
4.72
4.74
4.96
5.18
5.24
5.33
5.39
6.05
6.06
6.13

3.85
4.21
4.43
4.46
4.63
4.63
4.85
5.12
5.21
5.29
5.34
5.95
6.06
6.06

3.89
4.16
4.25
4.44
4.55
4.75
4.91
4.98
5.09
5.23
5.90
6.00
6.00

3.62
3.99
4.14
4.27
4.43
4.63
4.75
4.80
4.94
5.06
5.64
6.00
6.00

3.68
3.89
4.10
4.30
4.44
4.59
4.69
4.85
4.97
5.50
5.72
5.71

3.44
3.69
3.91
4.08
4.31
4.46
4.60
4.79
4.90
5.41
5.67
5.65

3.50
3.73
3.92
4.13
4.25
4.37
4.56
4.65
5.00
5.16
5.11

3.24
3.51
3.71
4.04
4.16
4.27
4.53
4.61
4.96
5.11
5.06

3.36
3.62
4.02
4.12
4.28
4.56
4.58
4.96
5.11
5.06

Maximum Trade Length

Maximum trade length (see Table 9) ranges from 13 days at various combinations of opening
limits above 240 and closing limits of 180 or greater to a high of 26 for an opening limit of 40. In
general, low opening and closing limits produce high maximum trade lengths. Closing limits
near the opening limit slightly decrease the maximum length. This suggests that high opening
limits signal situations where there is intense pressure to return toward zero while low opening
limits may capture situations where the current trend is away from zero.

10

Table 9
Maximum Trade Length (Days)
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
26
25
25
24
24
24
24
24
22
17
17
17
17
17
17
17
17
17
17

10
26
25
25
24
24
24
24
24
22
17
17
17
17
17
17
17
17
17
17

20
26
25
25
24
24
24
24
24
22
17
17
17
17
17
17
17
17
17
17

30
26
25
25
24
24
24
24
24
22
17
17
17
17
17
17
17
17
17
17

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

25
25
24
24
24
24
24
22
17
17
17
17
17
17
17
17
17
17

25
25
24
24
24
24
24
22
17
17
17
17
17
17
17
17
17
17

17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17

17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17

17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17

17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17

17
17
17
17
17
17
17
17
17
17
17
17
17
17
17

17
17
17
17
17
17
17
17
17
17
17
17
17
17
17

17
17
17
17
17
17
17
17
17
17
17
17
17
17

17
17
17
17
17
17
17
17
17
17
17
17
17
17

16
16
16
16
16
16
16
16
16
16
16
16
16

16
16
16
16
16
16
16
16
16
16
16
16
16

16
16
16
16
16
16
16
16
16
16
16
16

16
16
16
16
16
16
16
16
16
16
16
16

14
14
14
13
13
13
13
13
13
13
13

14
14
14
13
13
13
13
13
13
13
13

14
14
13
13
13
13
13
13
13
13

Maximum Length of Profitable Trades

The maximum length of profitable trades (see Table 10) ranges from 5 days at high opening and
closing limits to 9 days at combinations of low opening and closing limits. This parameter is, of
course, determined by some individual trade.
Table 10
Maximum Length of Profitable Trades (Days)
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
9
9
9
9
9
9
9
9
9
9
9
9
9
7
7
7
6
5
5

10
9
9
9
9
9
9
9
9
9
8
8
8
7
7
7
7
6
5
5

20
9
9
9
9
9
9
9
9
9
7
7
7
7
7
7
7
6
5
5

30
8
7
7
8
8
8
8
8
8
7
7
7
7
7
7
7
6
5
5

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

7
7
8
8
8
8
8
8
7
7
7
7
7
7
7
5
5
5

7
7
8
8
8
8
8
8
7
7
7
7
7
7
7
5
5
5

7
8
8
8
8
8
8
7
7
7
7
7
7
5
5
5
5

7
7
7
7
7
7
7
7
7
7
7
7
7
5
5
5
5

7
7
7
7
7
7
7
7
7
7
7
7
5
5
5
5

7
7
7
7
7
7
7
7
7
7
7
7
5
5
5
5

7
7
7
7
7
7
7
7
7
7
7
5
5
5
5

7
7
7
7
7
7
7
7
7
7
7
5
5
5
5

5
5
5
5
5
5
5
5
5
5
5
5
5
5

5
5
5
5
5
5
5
5
5
5
5
5
5
5

5
5
5
5
5
5
5
5
5
5
5
5
5

5
5
5
5
5
5
5
5
5
5
5
5
5

5
5
5
5
5
5
5
5
5
5
5
5

5
5
5
5
5
5
5
5
5
5
5
5

5
5
5
5
5
5
5
5
5
5
5

5
5
5
5
5
5
5
5
5
5
5

5
5
5
5
5
5
5
5
5
5

Minimum Length of Losing Trades

The minimum length of losing trades (see Table 11) ranges from 1 to 4 days. Highest values are
generally at opening limits of 140 to 280 and closing limits of 80 or less. As with the maximum
length of profitable trades, this parameter is potentially affected by some outlier trade. It may or
may not be profitable to truncate trade length in an attempt to avoid the worst trades.

11

Table 11
Minimum Length of Losing Trades (Days)
Rows show Opening Limits, Columns show Closing Limits
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
2
3
3
2
2
4
4
3
3
3
3
4
4
2
2
2
2
2
2

10
2
2
3
2
2
4
4
3
3
3
3
4
4
2
2
2
2
2
2

20
1
2
2
2
2
2
3
3
3
3
3
4
4
2
2
2
2
2
2

30
1
2
2
2
2
2
3
3
3
3
3
3
4
2
2
2
2
2
2

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

1
1
2
2
2
3
3
3
3
3
3
4
2
2
2
2
2
2

1
1
2
2
2
3
3
3
3
3
3
4
2
2
2
2
2
2

1
2
2
2
3
3
3
3
3
3
3
2
2
2
2
2
2

1
1
2
2
2
3
3
3
3
3
3
2
2
2
2
2
2

1
2
2
2
3
3
3
3
3
3
2
2
2
2
2
2

1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2

1
2
2
2
2
2
2
2
2
2
2
2
2
2
2

1
1
2
2
2
2
2
2
2
2
2
2
2
2
2

1
2
2
2
2
2
2
2
2
2
2
2
2
2

1
1
2
2
2
2
2
2
2
2
2
2
2
2

1
1
2
2
2
2
2
2
2
2
2
2
2

1
1
2
2
2
2
2
2
2
2
2
2
2

1
1
1
2
2
2
2
2
2
2
2
2

1
1
1
1
2
2
2
2
2
2
2
2

1
1
1
2
2
2
2
2
2
2
2

1
1
1
2
2
2
2
2
2
2
2

1
1
2
2
2
2
2
2
2
2

Time Segment Characteristics

Crush spread results are compiled for four equal non-overlapping time segments.3 See Table 12.
Results vary systematically over the time segments. The first segment had, on average, a
negative crush spread while the crush spread averaged 1,716 over the remaining three segments.
Although the fourth segment had the lowest standard deviation of the spread, it had the highest
kurtosis and the most positive skewness. This increased incidence of extreme changes potentially
leads to incorrect buy / sell signals. Trading results for the four time segments and for short
versus long trades are reported in the following section which includes transaction costs.
Table 12
Crush Spread
Time Segment Statistics
Segment

1
7/2/1984
Begin
12/1/1989
End
1,372
Observations
-36
Mean
Standard
Deviation
Skewness
Kurtosis

801
0.28
-0.45

2
3
4
12/4/1989 5/12/1995 10/18/2000
5/11/1995 10/17/2000
4/7/2006
1,372
1,372
1,372
1,784
1,372
1,993
1,066
0.43
-0.70

941
1.02
0.34

736
1.44
6.71

Transaction Costs
When transaction costs of $103.5 are deducted from each trade (see Simon for justification) the
percentage of profitable trades declines from the range of 72% - 80% to the range of 31% 74%.4 Additionally, the systematic difference between the length of winning and losing trades is
3

The four segments are 1 (July 2, 1984 Dec 1, 1989) 2 (Dec 4, 1989 May 11, 1995) 3 (May 12, 1995 Oct 17,
2000) 4 (Oct 18, 2000 Apr 7, 2006).
4
Transaction costs may be lower recently than during earlier parts of the study due to lower commissions. However,
the major portion of the transaction cost, bid-ask spreads, are unlikely to have changed significantly.

12

reduced. For example, at 380/140 the percentage of profitable trades drops from 80% to 71%,
while the average length of winning trades drops from 1.91 to 1.71 days, and the average length
of losing trades drops from 6.00 to 5.17 days. It is interesting that BOTH the (un-weighted)
overall average length of profitable trades (declines by .23 days) and the average length of losing
Table 13
Percentage of Profitable Trades
With Transaction Cost
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
41%
44%
48%
51%
52%
54%
54%
55%
55%
56%
60%
64%
65%
64%
67%
66%
71%
73%
73%

10
38%
41%
46%
49%
50%
54%
54%
56%
56%
58%
61%
64%
66%
64%
68%
67%
70%
73%
73%

20
36%
38%
43%
48%
49%
52%
53%
55%
55%
58%
61%
64%
64%
62%
67%
66%
70%
73%
73%

30
33%
35%
40%
44%
47%
50%
52%
54%
55%
56%
60%
63%
64%
62%
67%
66%
70%
73%
73%

40

50

60

70

80

90

100

110

120

130

140

150

160

33%
37%
42%
45%
48%
51%
54%
54%
55%
58%
62%
63%
62%
67%
65%
70%
73%
73%

31%
35%
40%
43%
46%
50%
53%
54%
55%
58%
62%
63%
62%
67%
66%
71%
74%
74%

33%
39%
42%
45%
49%
52%
53%
54%
57%
60%
62%
62%
67%
66%
69%
73%
73%

32%
37%
40%
43%
48%
51%
52%
54%
57%
60%
62%
61%
67%
66%
67%
72%
73%

35%
39%
42%
46%
49%
51%
53%
56%
59%
62%
61%
66%
65%
68%
72%
73%

34%
37%
40%
44%
47%
50%
53%
56%
60%
62%
62%
67%
66%
69%
72%
73%

36%
38%
43%
46%
50%
52%
55%
59%
62%
62%
67%
66%
70%
72%
73%

35%
37%
42%
45%
48%
51%
55%
59%
62%
61%
65%
64%
68%
70%
72%

36%
41%
44%
47%
50%
55%
59%
62%
62%
66%
65%
69%
71%
72%

35%
40%
43%
46%
49%
54%
58%
62%
62%
66%
66%
70%
71%
72%

39%
41%
46%
48%
53%
58%
63%
63%
67%
67%
70%
71%
72%

38%
41%
45%
46%
52%
57%
61%
61%
66%
66%
69%
73%
74%

39%
43%
44%
50%
55%
60%
61%
66%
64%
67%
73%
72%

trades (declines by 1.17 days) decrease as some of the longer formerly profitable trades become
relatively short losing trades. Consistent with this, the maximum length of profitable trades
decreases by 1 to a length of 3 days in most events at closing limits of 110 or less. Similarly, the
minimum loss length declines by 1 day at most combinations of limits.
Meanwhile, the coefficient of variation (see Table 14) of course rises, as profit has been reduced
by $103.5 while the standard deviation is unchanged. In the case of 380/140, the coefficient of
variation rises from 2.56 to 5.22. It is this change in coefficient of variation as a measure of the
risk-return relationship which is particularly troubling as explained below.
Table 14
Coefficient of Variation with Transaction Cost
Rows show Opening Limits, Columns show Closing Limits
0
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400

0
-5.69
-6.61
-7.83
-10.07
-12.96
-19.12
-90.88
65.69
29.95
13.90
9.16
6.79
5.50
5.16
4.50
4.97
4.75
4.64
4.99

10
20
30
40
50
60
70
-5.33
-4.88
-4.44
-6.13
-5.55
-5.07
-4.62
-4.25
-7.42
-6.66
-6.05
-5.49
-5.02
-4.58
-4.33
-9.61
-8.51
-7.80
-6.96
-6.33
-5.90
-5.55
-13.01 -11.06 -10.30
-9.15
-8.10
-7.44
-6.99
-19.74 -14.96 -13.89 -11.58 -10.32
-9.50
-8.80
-89.51 -37.68 -29.29 -23.44 -20.55 -18.33 -16.63
58.37 1637.49 -159.45 -62.80 -43.97 -40.70 -33.23
25.88
48.45
82.32 1009.99 -530.34 -226.15 -107.36
13.20
17.02
18.21
23.30
23.71
28.31
31.39
9.19
11.08
11.59
13.06
13.12
15.28
15.55
7.10
8.34
8.52
9.46
9.47
9.81
9.86
5.67
6.37
6.37
6.84
6.72
7.06
6.96
5.38
5.95
6.05
6.40
6.24
6.52
6.40
4.63
4.98
5.10
5.30
5.19
5.33
5.24
5.13
5.50
5.67
5.92
5.69
5.76
5.86
4.96
5.21
5.21
5.30
5.08
4.98
5.03
4.84
5.12
5.12
5.20
4.96
4.34
4.31
5.25
5.62
5.62
5.64
5.32
4.57
4.50

80

90

100

110

120

130

140

150

160

170

180

190

-5.24
-6.66
-8.48
-15.39
-28.60
-68.29
33.69
16.08
10.02
6.87
6.25
5.09
5.60
4.85
4.28
4.45

-4.99
-6.41
-8.28
-14.49
-25.86
-89.64
30.04
15.30
9.37
6.52
6.55
5.29
5.88
5.09
4.48
4.68

-6.00
-7.77
-13.13
-23.28
-57.77
36.53
16.88
9.70
6.71
6.62
5.27
5.85
5.07
4.53
4.75

-5.62
-7.20
-11.68
-18.44
-34.40
71.85
21.55
10.94
7.51
7.53
5.87
6.57
5.51
5.06
5.26

-7.12
-11.40
-18.26
-34.41
75.41
22.84
11.28
7.63
7.85
6.01
6.75
5.66
5.21
5.46

-6.77
-10.51
-15.98
-29.34
127.54
25.09
11.45
7.64
7.77
5.97
6.56
5.52
5.14
5.52

-10.54
-16.41
-31.06
85.69
25.19
11.77
7.83
7.87
6.10
6.70
5.59
5.22
5.60

-10.50
-16.50
-32.08
76.26
24.18
11.46
7.74
7.81
6.02
6.60
5.51
5.16
5.52

-12.01
-18.22
-83.29
55.27
14.17
9.09
8.75
6.57
7.37
6.04
5.48
5.87

-11.34
-16.71
-52.04
108.73
16.54
10.08
9.72
7.15
8.08
6.59
5.97
6.48

-18.04
-75.48
50.70
12.27
7.85
7.38
5.54
6.04
5.00
4.38
4.62

-14.92
-42.79
98.87
13.69
8.41
7.83
5.89
6.51
5.35
4.70
5.04

The coefficient of variation ranges from a low of 4.28 (380/80) to very high and then negative
values for those strategies with negative average profit. A coefficient of variation under 5 is
achieved in only 26 combinations of opening and closing limits, and all of these are at opening

13

limits above 300 and generally at closing limits of 100 or less. The exceptions are 4 cases with
limits of 380-400/180-200. The effect on the coefficient of variation is great. For example, the
average coefficient of variation for opening limits of 380 doubles from 2.46 to 4.90 with the
introduction of transaction costs. If the level of profit is normally distributed, which it is not, a
coefficient of variation of 4.28 (the best possible combination) would signify approximately a
55% probability of gain. The fact that the probability of gain is somewhat higher (72%) is a
result of skewness in the distribution of profit.5 These low probabilities of success are not likely
to attract arbitrage attempts.
Short Versus Long Trades
Results are compiled separately for short and long trades. See Table 15. Long trades are more
profitable on average and have a higher likelihood of profitability but have a higher standard
deviation of profit. Consistent with this are higher maximum gains and losses for long trades.
Table 15
Short versus Long Trades
Limits from 40/0 to 400/200

Average Number of trades

Average Profit
Average Standard Deviation of Profit
Minimum Coefficient of Variation
Maximum Gain
Minimum Loss
Average Maximum Trade Length / Actual
Average Percentage Profitable
Max Profitable Length (Average / Max.)
Min Loss Length (Average / Min.)
Average Profitable Trade Length
Average Losing Trade Length

Short

Long

169
$19.18
$337
380/70 4.09
$952
$2,893
15.83 / 26
53%
5.32 / 9
1.51 / 1
1.83
4.79

168
$47.97
$403
400/180 3.15
$1,210
$3,249
15.96 / 18
60%
4.79 / 7
1.68 / 1
1.87
4.63

Some combinations of limits do have a lower coefficient of variation for long trades. Statistics
for trade lengths are very similar for short and long trades.

Higher Limits
Higher opening limits (up to 800/400) are also studied with transaction costs. The number of
trades drops precipitously. While average profit fails to increase, the percentage of profitable
trades hovers in the 60-70% range. Probability of gain as high as 77% exists (560/0 and 560/40),
but the coefficient of variation is 5.95 and 7.24 respectively with average profit of $142.82 and
$115.23. In general the coefficient of variation at higher opening limits deteriorates, although it
5

At 380/140 skewness is -3.25 and kurtosis is 20.63. The effect of the skewness is not a constant. At low profit
levels the negative skewness acts to increase the probability of loss. For example, at 180/0 the probability of gain
falls from 75% before transaction costs to 55% with transaction costs. The coefficient of variation with transaction
costs is 65.69 which would indicate a 50.6% probability of gain in a normal distribution.

14

is as low as 2.24 at 800/400. Unfortunately only 8 trades exist over the 20-year study period at
this level and waiting for them is likely not justified by the $242 average profit.
Segmenting the Study
The four time segments are again analyzed for differences in trade results. In general, the fourth
time segment gives negative profit results, a higher standard deviation of profit, a lower
percentage of profitable trades, and longer trade length. See Table 16. These results demonstrate
that even should a profitable set of opening and closing limits be found, in general, specific time
periods may yield negative results. Traders need be aware of changing spread characteristics.
Table 16
Example Trade Results
By Time Segment
With Transaction Costs
Open/Close Number
of
100/0
Trades
1
231
2
181
3
194
4
199
200/0
1
116
2
49
3
85
4
90
300/0
1
63
2
13
3
41
4
31
400/0
1
30
2
5
3
21
4
14

Average
Profit

Percentage
Profitable

2
-18
-13
-66

Standard
Deviation
of Profit
277
154
265
320

57
50
49
46

Average
Trade
Length
4.01
3.82
4.28
4.38

31
26
74
-71

314
190
335
425

55
65
60
44

4.38
4.00
4.20
4.62

102
39
239
-98

336
191
386
693

65
62
66
58

4.13
4.15
3.41
4.81

112
83
403
-278

379
220
401
996

70
81
64
73

4.33
3.60
2.19
5.86

Reversal Strategy
The strategy of reversing trades only when the spread has reached a level equal but of opposite
sign to the opening limit (the approach employed by Simon) reduces the number of trades,
increases the average profit and standard deviation of returns, and increases the coefficient of
variation.

15

Table 17 examines the relationship between average profit, standard deviation of profit, and the
coefficient of variation over a variety of closing limits. In general, profits are maximized at limits
slightly beyond equilibrium but short of a complete reversal of sign. Opening limits of 100 are
clearly not profitable, regardless of closing limit, at transaction costs of $103.50 while profits are
maximized at limits of 200, 300 and 400 at closing limits of -100, -100 and 150 respectively.
Table 17
Profit, Standard Deviation of Profit, and Coefficient of Variation
Opening Limits (Rows) 100 to 400
Closing Limits (Columns) 200 to -400
Profit
100
200
300
400

200

150

100

43
85

-9
54
101

-5
63
114

Standard Deviation of Profit

200
150
100
100
200
300
400

351
442

289
416
553

Coefficient of Variation
200
150
100
200
300
400

8.16
5.20

-32.11
7.70
5.48

296
412
538

100
-59.20
6.54
4.72

50
-38
0
72
111

0
-26
11
88
119

-50
-18
17
107
139

-100
-11
20
110
171

-150

-200

-250

-300

-350

-400

-2
104
195

-20
63
148

48
100

143
162

89

80

50
241
329
443
590

0
265
342
452
595

-50
300
360
458
599

-100
357
401
475
609

-150

-200

-250

-300

-350

-400

517
556
617

604
634
675

814
855

1080
1099

1269

1393

50
-6.34
Undef.
6.15
5.32

0
-10.02
29.95
5.16
4.99

-50
-16.39
20.60
4.28
4.32

-100
-32.81
20.06
4.34
3.57

-150

-200

-250

-300

-350

-400

-206.95
5.36
3.17

-30.77
10.00
4.56

17.08
8.56

7.55
6.80

14.19

17.30

Simon, for comparison has the following results:

Opening
Closing
Avg.
Limit
Limit
Profit
Long Trades
-100
+100
27
-200
+200
231
-300
+300
355
Short Trades
+100
-100
33
+200
-200
185
+300
-300
185

Standard
Deviation
406
569
1,092
284
362
549

Some movement beyond equilibrium can safely be anticipated, though not complete reversal. At
the same time, the standard deviation of profit increases almost always as the closing limit moves
farther from the opening limit. This implies that the risk-return relationship (as measured by the
coefficient of variation) is optimized when closing trades only slightly beyond equilibrium as
indicated by the 5-day moving average.
While the standard deviations are similar, Simon experiences greater profitability over the earlier
portion of the longer time period employed in this study. As earlier discussed, the fourth time
segment reduces overall profitability.

16

Summary of Analysis
A risk-return tradeoff (coefficient of variation) indicates that contrary to the symmetric limits
studied by Simon (e.g. short the crush if the spread is greater than $100 above the 5-day moving
average and reverse the trade if the spread is $100 less than the 5-day moving average), traders
can benefit by reversing trades near the 5-day moving average. Average profit increases, the
standard deviation of profit decreases, and the percentage of winning trades increases.
Traders also may benefit from establishing trade length limits (i.e. truncating trades). Losing
trades are, on average, significantly longer than winning trades. Artificial trade length limits may
help to reduce the incidence and severity of the worst losing trades.
Truncation of Trade Length
Because the average length of profitable trades is lower than the average length of losing trades,
and because the maximum length of profitable trades is relatively short; it may be profitable to
truncate trade length. For example, no trade at any set of limits up to 400/200 produces a
profitable trade in excess of 9 days and almost no profitable trades over 7 days exist at opening
limits over 200. See Table 10. Truncating trades may or may not avoid losing trades, and could
exacerbate their losses if they are closed before a reversal signal is produced by the spread.
Three trade truncation limits are studied. In general, the coefficient of variation is improved as
the trade truncation limit is reduced to 10 or 8 days. At a truncation limit of 6 days the coefficient
of variation begins to increase. This is demonstrated with un-weighted averages of all
coefficients of variation at opening limits of 300 and higher. The region of 300 and higher is
chosen due to negative and extremely high coefficient of variations at lower opening limits
which indicate unlikely limits for arbitrage attempts. The average falls from 5.72 to 5.60 to 5.30
and then rises to 6.56 for unlimited trade length, 10 day, 8 day and 6 day limits respectively.
These same trade truncation limits have minimal effect on the probability of profitable trades.
The shortened trades avoid few losses and free up the trading account for additional trades in the
same unfavorable environment that caused the original trade to lose money.
Conclusion and Implications for Further Research
Unless further research demonstrates arbitrage limits (filters) with more appealing risk-return
characteristics, the soybean crush spread should be considered an efficient market. Although the
risk-return relationships are not consistent, they do not provide profitable arbitrage opportunities
using a variety of opening and closing limits.
In contrast to previous work by Simon, profitable trades are significantly shorter than losing
trades. However, truncating trade lengths does not significantly improve results.
The averaging period should be explored to ascertain if a longer (shorter) period better identifies
significant deviations. Similarly, other more sophisticated arbitrage identification techniques
such as Simons Fair Value, regression techniques, and neural networks may still yield
arbitrage opportunities.

17

The persistence of those arbitrage opportunities that do exist could also be explored. Further
research may want to examine the effects of doubling-up transactions i.e. creating second
transactions on top of existing trades. It would be interesting to note how long a trader has
available to exploit an arbitrage opportunity and the existence of such opportunities would say
something about market efficiency.
Bibliography
Barrett, W. B. and Kolb, R. W. (1995) Analysis of Spreads in Agricultural Futures, The
Journal of Futures Markets, 15, 69-86.
Butterworth, D. and Holmes, P., (2002) Inter-market Spread Trading: Evidence from U.K.
Index Futures Markets, Applied Financial Economics, 12, 783-791.
Carter, C. A. (1989) Arbitrage Opportunities between Thin and Liquid Futures Markets, The
Journal of Futures Markets, 4, 347-353.
Castelino, M.G. and Vora, A., (1984) Spread Volatility in Commodity Futures: The Length
Effect, The Journal of Futures Markets, 4, 39-46.
Cuny, C. J., (2006) Why Derivatives on Derivatives? The Case of Spread Futures, Journal of
Financial Intermediation, 15, 132-159.
Dunis, C. L., Laws, J., and Evans, B., (2006) Trading Futures Spreads: An Application of
Correlation and Threshold Filters, Applied Financial Economics, 16, 903-914.
Elfakhani, S. and Wionzak, R. J. (1997) Inter-market Spread Opportunities between Canadian
and American Agricultural Futures, International Review of Economics and Finance, 6, 361377.
Girma, P.H. and A.S. Paulson, (1999) Risk Arbitrage Opportunities in Petroleum Futures
Spreads, The Journal of Futures Markets, 19, 931-955.
Haigh, M.S. and Holt, M. T., (2002) Crack Spread Hedging: Accounting for Time-varying
Spillovers in the Energy Futures Markets, Journal of Applied Economics, 17, 269-280.
Johnson, R., Zulauf, C., Irwin, S., and Gerlow, M. (1991) The Soybean Complex Spread: An
Examination of Market Efficiency from the Viewpoint of the Production Process, The Journal
of Futures Markets, 11, 25-37.
Ma, C. K. and Soenen, L. A., (1988) Arbitrage Opportunities in Metal Futures Markets, The
Journal of Futures Markets, 8, 199-209.
Poitras, G., (1998) TED Tandems: Arbitrage Restrictions and the U.S. Treasury Bill/Eurodollar
Futures Spread, International Review of Economics and Finance, 7, 255-276.

18

Rechner, D. and Poitras, G., (1993) Putting on the Crush: Day Trading the Soybean Complex
Spread, The Journal of Futures Markets, 13, 61-75.
Simon, D. P., (1999) The Soybean Crush Spread: Empirical Evidence and Trading Strategies,
The Journal of Futures Markets, 19, 271-289.
Wahab, M., Cohn, R., and Lashgari, M. (1994) The Gold-Silver Spread: Integration,
Cointegration, Predictability, and Ex-Ante Arbitrage, The Journal of Futures Markets, 14, 709756.

19