Edge or market?
Available in the Edge or market? tab of the Performance window, in the Edge column of the Laboratory ranking and in rung 4 of AQ Genesis. It answers one question: does this strategy earn its profit with its entry, or is it a gift from the market?
Each strategy is compared with its «always» twin: the same strategy, with the same exits (TP, SL, bar exit, session, session close), the same entry order (market, Stop or Limit) and the same costs, but entering at any moment: as soon as it is flat, it enters again. The only thing that changes is the entry condition, which is exactly what we want to measure. In a rising index the twin makes money on its own: if the strategy does not improve on it, its filter adds nothing.
Everything is measured per day in the market
Comparing total profit is misleading: a strategy that is in the market 20 % of the time cannot earn as much as one that is in 100 %, and one that is in for a long time collects the market’s rise even if its entry adds nothing. That is why the unit is profit per day in the market, for the strategy and for its twin, year by year. The edge in dollars is that difference times the days the strategy was in. RoMaD (profit over drawdown) is not used either: in a rising index the drawdown is set by one big fall, and it would reward the strategy that happened to be out that day.
The measures
- Edge: what it earns above the twin on its same days in the market.
- Years it beats the market, with a sign test (p): winning 6 of 6 years is much harder by chance than 4 of 6.
- Without the best year: the edge after removing the year that contributes most. If almost all of the edge is one year (a lucky 2020), it is not an edge: it is an episode.
- Percentile vs random: the result against 200 twins that enter at random with the same days in the market, year by year. Shown for information; it does not decide yet.
- Time in the market of the strategy.
The traffic light
It earns more per day in the market than the market itself in at least 60 % of the years (sign test p < 0.10) and does not depend on a single year.
It wins more years than not, but it is not conclusive yet. It passes the filters, with the warning in sight.
It beats the market half of the years or fewer, or removing its best year leaves less than 10 % of the edge. Its profit comes from the market.
If a period covers less than 3 months of its first or last year, that piece is added to the neighbouring year: a cut in November does not turn six weeks into «a year». The thresholds were checked on the 100 strategies sealed by Genesis on daily @ES: out of sample, the red ones beat their twin only 35 % of the time; the green ones, 77 %.
In the Performance window
One card per period (in-sample, out-of-sample and full history) with its traffic light and a sentence that explains it. Below, the edge bars per year, with the cut line, and the curve of what it earned against what the market would have paid on its same days in. Both are computed with 1 contract, without Position Size: what is measured is the entry.
In the Laboratory
When the sweep finishes, the whole ranking is measured (the twin is computed once per combination of exits). The Edge column has two dots: in-sample and out-of-sample; hover to see the detail. Click its title to sort, and the Hide market button hides the red rows without deleting anything. If you want more to choose from, raise «Up to N results» in the sweep settings.
In AQ Genesis
It is rung 4, «No edge vs market». It discards the red ones in training (in the cheap filter) and in validation (in the wall, before Walk-Forward). In the OOS reserve it only informs: the reserve never decides. Each sealed strategy shows three dots (training · validation · reserve): in Genesis New, the Edge column of the Summary and Wall views; in the classic window, next to the name.
What it does not do
It separates edge from market, not from luck. In-sample, almost any winner of a sweep comes out green: the sweep picked it precisely for beating the market there. And in a shuffled market, with no trend, the twin earns nothing and almost any selected strategy beats it. A red strategy only collects what the market gives; a green one is not necessarily good. For luck, rely on out-of-sample, PBO, the wall and the shuffled market.
A real case. A long strategy on 5-minute @ES made $185,000 in its search period (2016-2021) with a RoMaD of 8.54, against 2.06 for its twin. Measured per day in the market, almost all of its edge was 2020: without that year, $1,372 of $75,714 remained. Red. Out of sample, it earned less than entering at any moment.
AniQuant can be tried free for 14 days, with every module and no card. If you have the book, 30.