⏱️ AQ Holding Profile how long to hold
It opens from Market → AQ Holding Profile…. The same idea as the TP/SL Map —it enters on EVERY bar, with no filter and no signal, so that what gets measured is the market and not a strategy— but changing the axis: instead of sweeping TP × SL in percent, it sweeps the time invested. It answers a question the rest of the program had no way of answering: how long is it worth staying in?
The main curve: how often am I right?
For each holding length N —from 1 to the maximum you choose, all of them at once— it enters on every bar of the stretch, exits exactly N bars later and records the result. The first thing you will see is the most direct one: if you hold for one day on @ES you close in profit 53.5% of the time; if you hold for sixty, 66.6%.
Beside it runs its reference line, dotted: the probability that drift alone would explain, given the mean and the volatility of that same return. The shaded gap between the two is what is not drift — and there something shows up that you cannot see any other way:
Observed 66.6% · drift alone 59.2% → +7.3 points. Index trades are profitable more often than their mean and volatility predict: there are many small gains and few large losses, so the median sits above the mean.
Observed 60.9% · drift alone 60.6% → +0.3 points. Gold is profitable exactly as often as it should be. The same question, two different answers: which is why the reference is needed.
Why that figure on its own misleads
The «% of times it wins» always rises with the holding length in a market with drift. If the window showed only that, it would answer «hold as long as you can» on every rising instrument in the world. Which is why two counterweights sit beside it, and the answer is the three of them together.
The total return divided by the holding length. On @ES it comes out flat —0.0209% ± 0.0003 from N=1 to N=250—: holding does not pay more per unit of time, it only accumulates. On @GC it rises to 0.0240% at N=120: there is a preferred holding length there.
How far the trade goes against you —and in your favor— along the way before you get out. It is the price paid in suffering: on @ES, −0.57% at one day against −4.42% at sixty. Fifteen times the adverse excursion for the same performance per bar.
The window marks the maximum of the quality per unit of time (return ÷ deviation, normalized by the holding length) and calls it the preferred holding length. And when that curve is flat it says «none» instead of pointing at the peak of some noise — which is what happens on @ES, and is as useful a piece of information as the number itself.
What else is in the window
One's return is the other's with the sign flipped, but the MAE is not: a long suffers the falls and a short the rises, and in a market with drift they do not measure the same thing. The switch at the top changes side without recomputing anything.
The entries go between the two; the exit may indeed fall after the second one, so the trades at the end of the stretch are resolved with whatever bars come next instead of being lost. It is the TP/SL Map's same «cushion». It is what you use to ask «and over the last five years?».
Every holding length with its eight columns, sortable by any of them and navigable with the arrow keys. The chosen row is marked with a vertical line on all three charts, so you can read one specific holding length against the three questions at once.
At the top, in one sentence, the three things together: how often you are right, whether that is more than drift explains, whether you get paid more per bar, and at the cost of how much adverse excursion. Any one of them on its own always misleads.
All three charts have a magnifier. The engine skips the bars with a close ≤ 0: this is not theoretical — a back-adjusted continuous like @CL can have them, and there any calculation in percent is garbage.
It is not a strategy, nor a backtest: there is no entry signal, no stop, no commissions and no slippage, and the trades overlap because it enters on every bar. It is a portrait of the instrument, not a result you could obtain. It is there to help you choose, with some judgement, the horizon of the other tools —the Pattern Miner's H, the Edge Scanner's horizon, a time stop— and to know what you would have to beat for your entry to add anything.
AniQuant can be tried free for 30 days, with every module and no card.