I like back-testing or paper-trading various strategies and adjustments, but I don't trust the results to be replicated in live trading situations. Last week, I discussed some of the difficulties in translating free-wheeling adjustment ideas into a live trading situation. Those who would like to review the article can find it here.

The specific situation discussed was a 50-point RUT NOV11 10/20/10 put butterfly at the 760/710/660 strike with an additional NOV11 770 call as a delta hedge, established 30 minutes before the close on October 14. By October 27, the RUT's price had moved to and then outside the expiration breakeven, one adjustment point for many butterfly traders. The long call would have hedged the butterfly's losses somewhat.

A brief review of last week's considered adjustments might be in line for those who don't want to go back and review the original article. In back-testing, a trader might have elected to adjust by adding another butterfly centered at a point a little ahead of the RUT as it was hitting that expiration BE. When back-testing, it would have been easy to add that butterfly. The conscientious back-testing trader might even have added an additional cost to the trade to account for slippage in live trading. However, if the new butterfly had stepped on some of the strikes of the original butterfly, it might not have been possible to execute the trade on some platforms. Depending on the platform, then, and the strikes chosen, the trade might have needed to be broken into two or more trades, adding more opportunities for slippage. Back-testing and reality might not have matched.

An easy solution existed for that problem, however. If the original butterfly was an all-put butterfly, the new one could have been placed as an all-call butterfly, all in one trade, eliminating the additional risks associated with breaking a trade up into several legs. In many market environments, it would have been possible to execute the adjustment as a single trade, although that is not always true of fast market conditions.

However, what if the trader did not want to add an additional butterfly and increase the margin or buying-power effect that much? What if the trader preferred to lift up the entire butterfly and place a new one? Again, when we're back-testing, that's easy to do. We just set up the orders. We can even build in some slippage if we're trying to be conscientious. However, in this case, particularly in market conditions like those incurred on October 27, it's likely we're going to incur some time-related risk, too.

For example, I like to avoid adjusting in the first half hour to hour, if I can, because prices tend to be inflated and sometimes the first move is reversed. However, on October 27, the close of the first hour of trading found the RUT's prices pressed up against the butterfly's expiration breakeven that was then at about 746.58. Using the proprietary PnL charting program that I use, I keyed in the adjustments needed to lift up the original 760/710/660 put butterfly and buy a new, re-centered one at 800/750/700. I could click a button to commit the trade, and I would have a new trade with a total cost or $10,307, including two-way commissions. I have new delta of 5.55, with a gamma of -1.43, a theta of 100.88, and a vega of -154.31.

But . . . what if it I sell the old butterfly and only then enter the order to buy the new one, as would often happen in real life? (Or, do the opposite, as some butterfly traders do, buying the new one and only then selling the old one?) And what if there are a few minutes between the time I could execute the first trade and the second one, as often happens, particularly if I have to plumb for the executable price? My program tells me that with five minutes between the two trades, the total cost of the trade would now be lower, at $10,217. Delta is 6.08, gamma is -1.41, theta 101.24, and vega -153.10. Add another five minutes and the total cost is now $10,442.00. Wait thirty minutes and the cost has sunk to $9,842.00.

What if I had waited until 90 minutes after the open to close the first butterfly? Back-testing opening and closing both trades at once would have resulted in a total cost of $10,007.00. Opening the new butterfly five minutes later would have resulted in a cost of $10,067.

Obviously, the cost differences shown here are not huge, but they do exist and they can be much larger in live trading situations. The differences would have been exacerbated if I had had to sell an existing butterfly below the mark and buy a new one above the mark. They might have been even more exacerbated if I had bought back the original butterfly in a quiet moment or during a brief pullback before the RUT shot higher again. Some traders might elect to stabilize the trade in a runaway upside or downside move by buying long calls or call debit spreads or long puts or put debit spreads while they get the trade executed, but that presents another opportunity for slippage. Some might elect to take off the original butterfly and wait until prices settle, perhaps even on another day, to re-center the new trade. I know of several traders who have elected this route lately when they could execute one part of their butterfly adjustment but not the other.

Each of these tactics had pros and cons, but one thing is certain: even if I had built in slippage into my back-testing, it wouldn’t have accounted for these types of situations.

I believe I learn a lot from back-testing or following simulated trades. They show me that some strategies and adjustments are going to be losers, no matter what I do. They show me that some others might prove workable. They give me confidence enough to trade, but they don't seduce me into believing that I'm going to go eleven months in a trade and never have a loss.

I have had trade strategies that went even longer than eleven months without a loss, but I have to see it in live trading before I believe it.