but to get back to opie, In paper, he sees a trade and takes it, in real life he waits for "conviction." The funniest thing was when my paper got way ahead of my real and rather than wait to reset it I tried to lose enough to get it back to even with real. I did everything I preach against, trading eur.usd, and I couldn't lose money to save my soul.
Yes, your trading style is what dictates the fidelity at which you need to model fills in simulation. I trade algorithmically and frequently, so I've built a simulation framework that models queue position (using the book) and various latencies . My principle assumption is my market impact is negligible...
There actually might be a real reason for why this is occuring. Can u tell us your time frame and method of entering orders? Which simulator are u using?
It really depends what you are trading and what simulator you are using. If you are trading stocks then there are many tricks being played by HFT. They can get in to an order before you by deceasing the price less than a penny. They have the ability to place their servers right at the stock exchanges. Plus they pay to see order flow and can react a less than 1 second. Now it could also be the simulator and data feed. Both can be off. However using a good sim with a good feed and instruments that allow you to trade within the bid and ask using either limit or market orders solves this problem for the most part.
Some simulators are better than others. For example, let's consider the fills obtained using the TOS paper trade platform (TD Ameritrade). This is a pretty good simlulator as it uses real market data, but it has one flaw that could be disastrous for novice traders. The TOS simulator's fill engine gives highly misleading fills. Normally, traders using a retail broker like TDA and the regular TOS platform may not get placed particularly advantageously in the order queue. What this means in practical terms is that if you place a limit buy order, a "resting" order, you can expect to be far enough down in the order queue that your chance of being crossed with a market sell order isn't all that great, consequently price will likely have to drop through your bid by a tic so the inside offer (ask) can be crossed with your bid, and you can get filled. The opposite situation pertains when you have placed a limit sell order. Effectively, what this means is that you are usually going to be buying on the inside ask (or offer) and selling on the inside bid. Now when you use the TOS paper trading platform a miracle takes place. Buy orders get filled everytime on the inside bid instead of the inside ask. Similarly, if you have a resting sell order in place, it will get filled as soon as the market's ask price (i.e., the inside ask) rises to your ask price. This way of filling orders will produce two tics in your favor on every round trip. You can scalp all day long and get rich very fast on paper, just like you were a market maker, buying consistently on the bid and selling on the ask!. I wrote TOS a long time ago about this issue, suggesting that they change their paper platform fill engine to fill buy limit orders only after the inside ask matches your bid, or the inside bid matches your offer if your selling. Which, although still an imperfect fill engine, would at least give fills much closer to reality.. After writing more than once, I eventually got a polite, perfunctory, but not very satisfactory from my point of view, response from Don Kaufman. Essentially, it amounted to a bit of arm waving. In my personal opinion, it is shameful for TOS to continue filling paper orders this way because it gives the neophyte a very false picture of how easy it is to make money trading. (I think we can safely assume we know the reason why they don't correct this fill engine, which would be trivial to do.) If you are using the TOS paper trade platform, be sure and subtract two tics from any profitable trade and increase your loss by two tics on any losing trade. Also pay attention to see if the market's price drops through (rises above) your bid (ask) by at least one tic. If it doesn't you likely would not have got a fill in actual trading even though you got a fill on paper!
This means that their simulator is not "pretty good" it is worse than useless and extraordinarily misleading. If they are aware of the issue (as they must be) it means that it is deliberately misleading would-be traders into thinking they might make a profit with their short-term trades. It is not just about subtracting "two ticks" it completely ruins the validity of any testing that you do. The problem you get in real life is not just fewer fills on the winners, you get adverse selection, so you end up with big losses (as the original poster stated) because as you are late in the que, u only get a fill when the market is going to keep running against you, on average. HFT's can actually identify retail order flow and trade against it, knowing that it is prone to losing over a short duration of time.
You can figure it out yourself. Put the trade on both paper and real money accounts and then see whether the real money trade goes against you while the sim one is profitable?
I totally agree that the TOS simulated fills are just bad and it's well known. But when it comes to non-paper I pretty much disagree that by using TDA the fills are any different. The fills are pretty much exactly the same as anywhere else once the limit order is placed. I use IB for 99% of my futures trading but I have screwed around with the book (not manipulation or anything) from my TDA account at the same time and saw no latency or other goofiness when submitting orders. Now granted, while I obviously could not see queue depth, I could definitely see from my IB account my TDA orders being added and removed from any given level with no perceptive delay from either TOS or the mobile app on a phone. Everyone submits their orders to the exchange in the same way - it's not like TDA is going to go out of it's way to internalize or otherwise screw with futures orders. For 99% of cases the broker doesn't really matter for this type of stuff, it's merely pre-positioning or reaction time (ignoring non-changable things like network latency) that will get you a good queue position. I would not doubt that they're capable of roughly discerning retail orders from the tape or potentially some form of crude detection of change in size at a given depth - but there's no reasonable way HFT is going to know specifically where retail orders are relative to non-retail within the queue itself. They may have a good idea of retail price points and typical reactionary levels but your fill success isn't going to be based on an HFT algo knowing you're a retail order in the queue. Things like obvious retail stop zones are not part of the queue they're part of knowing where retail traders typically put their trades on.
I agree that when using the regular TDA platform the fills are essentially what you will get with just about any, regular retail broker. I hope I didn't imply otherwise! In general you will be buying on the ask and selling on the bid. It's the paper trade platform that turns this around giving you two extra tics you wouldn't ordinarily get, and giving you a fill even if price does not drop/rise thru your order. That's why you must check every order executed using the paper plarform to see if price just reached you bid/ask or traded through.