I'm looking to grab 10 to 20 pt moves on NQ whereas before I've targeted more like 40 to 50 pt moves. Problem I'm facing is how do you accurately backtest a strategy that can have you out in a minute? I use Thinkorswim where you can only go back 30 days on 1m candles, so when I backtest scalpy trades on the 5m chart, I know there's a lot of error baked into the results because you can't always tell with 5m candles whether they'd stop you out internally or not. I found this out the hard way after doing a 6 month backtest on 5m candles, then seeing the edge disappear as soon as I got to the 30 days where I could check it against the 1m chart, and you see how much more noise there is within a 5m candle. Is the only solution patient foretesting then? I generally don't like to even try a strategy with real money unless my backtest goes back 6 months to a year, but I don't feel like waiting that long to get scalpier. What praytell should I do?
I did that a few years ago. - You need 1m data, tick data is useless because you are never going to be that fast. - The problem is at filling your trades, both at enter and exit. It is very easy to backtest it when you assume perfect fills, what is the most common way of approach it. If you see backtests on futures out there they always asume perfect fills, that is why you need to write your own backtester. Focus on getting a realistic fill on your trades, as in some of the trades won't get filled until the price is against you. You could bypass this issue by pretending to use market orders that fill 5 ticks away from the midpoint, but if you are scalping that is not going to give you what you want. Filling limit orders when the price is against you is the way to go, because that will mimic an exchange. You can also use products that are ready for having an exchange locally like: https://www.alliedtesting.com/exchange-simulator-sim/
Move to TradeStation which uses LIBB = tick-by-tick LookInsideBarBacktesting I'm sure there are other, likely more expensive alternatives, but it is the only way to truly know if your backtest is anywhere near reality. I shake my head anytime I read that someone has done backtesting on anything above tick-by-tick and found their strat profitable. Sure it is!!!!!!!
Gotta love it when some QA monkeys are trying to sell a market simulation product. The whole thing reads as pure LOL. You are essentially testing a market-making strategy, there is a garden variety of approaches for it and a lot of possible issues. Ultimately, you can actually simulate order queue using MBO or at least keep track of size changes ahead and behind your own queue position (the latter requires assumptions about cancellation rates ahead and behind).
Scalp trades are not 10-20pt moves... I scalp 4-5 ticks on ES. Buy data is way to go, using tick data.
Who do you buy your data from? I'm trying to get more savvy about this aspect of things, but I'm not great with technology.