Oh, good, it seems we have yet another chance to review a trading system...only if they want to cooperate. I want to thank everyone for their input in the "chat room" thread, and I have added testing to "picks" etc., (thank you Brandon), and hope to post some results. I will not post anything if I don't have permission from the owners of the sites. I am always a bit cynical, but I would be glad to see some real numbers using real money from this S&P system. I recently had one of our traders (at near retirement age) get involved in something like this, and am fearful that he may be getting taken advantage of, I hope not. Anyway, if someone is using this system, or if the owner wants to share info, I would be glad to add it to the mix.
that's not necessarily true. system's ability to over-optimize is a function of a) the number of trades and b) the number of system parameters (ie degrees of freedom). therefore, by adding parameters, it is perfectly possible to over-optimize a system having any number of historical trades. - jaan
John, a naive question maybe, but always wanted to ask a system vendor this question in person. If your stuff is so good why do you bother with running a business selling it, rather than just trading it?
DT-wav - The bid/ask spread on the S&P emini is usually 0.25 (on the NDX emini it's usually between 0.50-1.00) and there is usually more than enough volume on either the bid or ask to satisfy an immediate 5 S&P emini contract fill (the equivalent to one large S&P contract). So IMO there's no real reason to deal with the potential slippage and delays of pit fills - just trade the equivalent number of eminis electronically.
ArchAngel - I know, spread on emini SP is often 0.25. There's often more than 5 contracts on bid/ask. Sometimes more than a 100. I can see this in Livecharts. I ask about big SP contracts spread. Livecharts SP01Z quotes are a total mess. You say that executions are better on Eminis vs Full contracts. Why all systems are based on Full SP contracts? Slippage - if you trade with less than 15 Emini contracts - very often you can get execution at one price by sending market order. Trading larger sizes requires different strategies. Multiple orders? One big market order? DT-waw
"If you are not the Todd B. that I thought I appologize....... just the way you spew out STENDAHL outlier terminology and Larry Stein from ALARON phrases from THE R&W technical services days makes me think that way......" . The above is an example of GobbledyGook at its finest. This guy definately knows what he's talking about. LOL!
DT-Wav - You probably see "system" results based on full SP contracts because it'll inflate their numbers by a factor of 5. It's marketing. But there can be some substantial slippage trading the full contracts that most of these system results don't even really take into consideration. They throw some number in there, but it's not necessarily a good representation. You could easily get 1/2-1 point slippage thanks to the shenanigans in pit executions, which means slippage could be as much as $125-$250/contract. And if there are fairly fast breaks in the price action, forget it - the system results will be way off from reality. I'm looking at the SP emini right now and it's currently bouncing around in the 50-100 contract range on both the bid and ask with a 0.25 spread. So you could pretty easily trade 5-10 full contract equivalents on the emini without problem. Re: Livecharts - first off, it's not a trader's data feed solution. It's more for casual investor types who want to check on things. The whole technology plant supporting it is fairly primative. If you're putting up real money to trade, you should at least step up to the QCharts datafeed instead of using their Livecharts - what are you saving by using Livecharts anyway, maybe one e-mini point a month? Not worth it. Also, note that the full contract data isn't as reliable as the e-mini data (the prices are overheard from the outcry in the pit and someone keys them onto the tape). You don't get bid/ask sizes and you won't necessarily get all trades or a completely accurate bid/ask quote. The emini data comes out of the electronic Globex system so it'll be more accurate than the pit data.
Why all systems are based on Full SP contracts? Two potential reasons. If they are showing you backtest data for more than 3 years, emini data only goes 4 years, and wasn't immediately liquid when it was first introduced in 1997. Also, because the tick size on S&P's is only 0.10, most systems backtest much better with a smaller tick size. If you buy on high + tick, and sell on low - tick, your results will be better by 0.30 per trade. If the average trade on the day trading system is in the 1.00-1.50 range, that extra 0.30 makes a huge difference. Note that the longer the time frame, the smaller the impact of tick size on backtesting system results. Anytime you see a system vendor show you an S&P system based on full SP contracts, ask to see the results on emini as well. In my experience, if you're dealing with 10m bars and longer the results should be similar. If you're dealing with 5m bars or less, there will be some difference in the results depending on the entry & exit techniques. Between 5m-10m your mileage may vary. Make sure slippage numbers are realistic, and take a look at the largest 10% winners and largest 10% losers and check the price bars on entry & exit. Were there wide range bars involved in the entry or exit? If so, slippage may be a factor.
I looked over the information on the Gunslinger website, referenced in the first post on this thread. The backtest results are very impressive for a daytrade only system. Of course, there is no way to verify the results or to be sure they were not produced or enhanced by curve fitting, but the large number of trades, over 700, would seem to argue against curve fitting being a significant factor. The system had a max drawdown of 49%, which it apparently produced in its first month of operation. I suppose it says something for the vendors that they didn't just test from 2/97 onward and avoid that drawdown, but that is a pretty big drawdown. I doubt too many people would keep trading a blackbox system that destroyed half their account the first month. I also wondered why they only tested from 1/97 onward. Why not go back to the inception of the contract, as tick level data is available for far longer than 1/97 onward. Most likely the answer has to do with wanting to produce big numbers from the bigger daily range in later years. Also, the range-bound period of "92-'94 was a death zone for S&P systems. Finally, I think the $70 slippage and commission allowance is far too small, particularly for a system that trades frequently. $15 is a standard RT commission, so that leaves a whole $55 for slippage on both sides of the trade. Sorry, not realistic unless the entries are on limit orders. More likely is giving up .5 on each side, which means slippage of around $250 per trade. Multiply 776 trades by $200 extra slippage and you get the astounding sum of $155,200. Ouch. I'm not picking on this vendor or suggesting there is any thing wrong with him or his system, pretty much all of them do it this way. I just can't imagine actually buying a system or risking money on it without knowing the logic or having had an independent party test it.