The order a list opens in is a recommendation, whether or not anyone intended it as one. You will look at the top few rows, form an impression, and screen the rest against that impression. This is not a character flaw, it is how everyone uses a sorted table, and it means the ordering you land on is doing more work in your decision than any single column you later read carefully.
The Momentum Trading Engine gives you real control over that ordering, and the control is worth using deliberately rather than accepting whatever comes up. The leaderboard tab exposes a rank-by selector with six choices, Overall Score, Total Return, Risk-Adjusted Return, Win Rate, Alpha and Profit Factor, alongside a scope filter for all time, crypto and stocks. Six orderings of the same set of strategies. Five of them are going to give you a different first row.
The number the page leads with sets the frame
Before you touch a sort control, the page has already framed the problem. The strategies tab opens with a header line reading nine strategies, best return plus 430 percent, average win 96.59 percent and average Sharpe 1.58, sitting above six tiles covering strategy count, backtested trades, win rate, drawdown, run-up and profit factor.
Best return of plus 430 percent is the largest and most emotionally legible number on the screen. It is also the single worst basis for a shortlist, because a best-of figure is an extreme value out of nine attempts, and extreme values are the observations most contaminated by luck. When a page leads with its maximum, every strategy you look at afterwards gets judged against a number that one strategy produced once.
Why a return-first ordering hurts you specifically
Sorting by total return puts the strategies with the most extreme outcomes on top, and extreme outcomes come from two sources that look identical in a table. One is a genuinely good rule set. The other is a strategy that happened to run on an instrument that went vertical during the test window.
The engine gives you a way to separate them, which is that Alpha and Total Return are offered as separate rankings. A strategy near the top of both lists earned its return. A strategy near the top of the return list and nowhere on the alpha list was carried by its instrument. Sorting by return alone gives you no way to tell, and the return sort is the one you will reach for by reflex.
There is a second problem, which is that return is the column most sensitive to how long a strategy has been running. A strategy tested over three years and one tested over eight months are not comparable on cumulative return, and a sort by return silently rewards the older one. Risk-adjusted return is at least partly normalised for that, which is one of several reasons it belongs earlier in your process.
The sequence I would actually run
Four passes, in this order. The order matters more than any individual pass.
- Filter before you sort. The engine lets you filter by asset class and by plan tier. Cut to the asset classes you can actually trade at your broker, and cut anything on a timeframe you will not be awake for. A 30 minute strategy on a US equity is a strategy you will miss half the signals on if you have a job, and no sort order fixes that.
- Rank by Risk-Adjusted Return, not Total Return. Take the top eight names. This is your candidate pool and you will not add to it later.
- Re-rank the same set by Alpha and by Profit Factor and note which names hold their position. A strategy that stays in the top half across all three is a different animal from one that tops a single ranking and falls to the bottom of the others.
- Rank by Win Rate last, and use it as a disqualifier rather than a selector. In this engine the average win rate is already 96.6 percent, so a high win rate distinguishes nothing. What it can do is flag the strategies at the very top of that range as candidates for exit rules that bank small gains and hold losses.
What the shortlist is missing that no sort will show
Every ranking on offer is computed from the same backtest, so no amount of re-sorting adds information. The engine reports 350 trades backtested across all strategies, which across nine strategies is roughly 39 trades each. Every one of those six rankings is a different summary of the same 39 numbers, and reordering them does not make the sample larger.
This is the honest limit of screening. Sorting well protects you from picking the worst candidates. It cannot protect you from a library where every candidate rests on a short history. That is what the trade log is for, and it is the tab most people skip. It separates live rows from reconstructed ones, and at capture it carried an explicit banner stating that 44 of its rows had been reconstructed by replaying the strategy over historical candles, with no order placed and no fill occurring.
Those 44 rows are exactly as informative as a backtest, because that is what they are. Before you deploy anything off your shortlist, open the log filtered to that strategy and count how many rows are real fills. If the answer is a handful, you are not selecting between proven strategies, you are selecting between hypotheses, and the correct position size for a hypothesis is smaller than the one for a track record.
The pass that costs you nothing and changes the list
Here is the thing to do this week. Run your normal screen, whatever it is, and write down the three strategies you would have deployed. Then rank by Risk-Adjusted Return and by Alpha and write down those top threes. If your original picks do not appear on either list, you were sorting by return and calling it research.
Then take whichever names survive and check one thing the rankings never surface, which is whether two of them are the same trade. With nine strategies built from eleven entry systems that the engine groups into momentum, breakout and mean-reversion, and five synchronised timeframes underneath them, two top-ranked strategies can easily be the same idea running on different bars. Deploying both feels like diversification and is not. Sorting cannot see that, and it will happily hand you three rows that go long the same thing on the same day.