The hardest part of trading isn't finding a strategy. It's following one.
We built AlgoTrader NQGC the other way round: rules set before the trade, five fixed tests before anything is offered, 87 strategies tested, 65 rejected, every forward-tracked trade shown, losses included.
Agree that most bias hides in how the signal is built, not in the rule on top. The half that gets skipped is selection: test enough MA lengths and the best one looks good partly because it was picked. When you test these, do you keep a count of the variants tried before choosing one?
@btgraveyard@TakeProfitLLC Thanks, that covers it. Being flat by the window end keeps the cost picture honest too: no overnight funding or gaps left to argue about.
@SJosephBurns Writing size down before the open is how we deal with that one: the excitement still turns up, it just has no lever left to pull. Do you fix size per setup ahead of time, or cap it by the day?
@quantscience_ Kelly sizing is only as good as the edge estimate you feed it. Overstate that a little and full Kelly becomes a very efficient way to meet your drawdown, which is why the forward track is where the estimate really gets graded.
@tradedispensary Impatience might be the most expensive input in trading. Drop a timeframe and the cost per trade stays put while the average move shrinks, so the signal has to work harder just to break even.
A hypothetical backtest found a statistically strong signal on micro Nasdaq futures. Then we counted the commission.
A strong signal on paper can still be too small to pay its way.
@SJosephBurns Agreed, and a lot of that 9:25 freeze comes from still deciding at 9:25. Rules written the night before turn the click into a check rather than a decision.
@tradedispensary That fits: a longer hold lets the average move grow while the cost per trade stays put. Did the signal keep its strength on the higher timeframe, or did it need reworking?
@quantscience_ Regime labels are easy to fit after the fact. The harder half is how many bars a model needs before it notices the regime has changed, because that lag is where a strategy pays for it.