Study · October 12, 2026 · updated October 14, 2026 · 11 min read
Momentum Strategies Backtested: 49,000 Trades After Costs
FyndStock editorial team · How we work
We tested the five rule sets of the FyndStock screener on about 49,000 simulated trades from May 2022 to September 2026, with stops treated as real orders and costs of 0.2% per buy and per sell. After costs, three strategies are profitable and two lose money. Against the fairest benchmark — a random purchase with the same holding period — only the Zanger strategy keeps a clear edge. Our first backtest had shown much better numbers. They were wrong.
This article exists because we were wrong. We first checked the screener's exit rules with a backtest and got profit factors between about 1.6 and 2.5. That looked excellent. When we set out to publish the numbers on our strategy pages, we set up the calculation more strictly once more. Afterwards, little of those nice values was left.
We show both here: what went wrong and what is left after an honest calculation. Most backtests you'll find online make the same mistakes our first one did. Once you know them, you read numbers like that differently.
How we tested
The basis is daily prices for 887 US stocks from the S&P 500 and S&P 400, going back six years. The test window runs from mid-May 2022 to mid-September 2026; before that the indicators need warm-up. A trade happens when a stock newly rises into the screener's top levels (ELITE or STRONG). For that, at least four of the five rule sets have to agree, the stock has to be in an intact uptrend, its sector has to be above average and the S&P 500 has to be above its 200-day average. That's how the live screener works too — the test calls its code directly.
Each strategy that confirms such a stock then trades it with its own rules: Minervini with a stop below the 50-day line, Zanger with 7.5% and Confluence with 8% distance, Qullamaggie and Kell with stops based on the average daily range (1.5 and 1 ATR). On top of that comes a target as a multiple of the risk and, in this first part, a maximum holding period of 7 to 25 trading days. Because the entries are largely the same for all strategies, this test mainly compares one thing: how you get out of a good candidate again.
Our first backtest was too optimistic
The mistake was inconspicuous. The first test checked every evening whether the close was below the stop and then booked the trade at the stop price. If a stock with a stop at $100 closed at $96, the test recorded a sale at $100. No broker in the world fills like that. If a stock gaps down far below your stop after bad news, you sell at the open, not at your preferred price. The first test also had no costs at all.
The new calculation treats stop and target like real orders resting in the market. It checks against the daily high and low, fills at the open on gaps and takes the stop first when both were touched on the same day. Every buy and every sell costs 0.2% for commissions and spread. The result:
Profit factor per strategy: first backtest vs. realistic calculation
Profit factor = sum of gains divided by sum of losses. Below 1 the strategy loses money. Realistic = stop as a real order against the daily range, gaps filled at the open, 0.2% costs per buy and per sell.
Qullamaggie is hit hardest: from 2.54 to 0.88. His stops are tight, so he gets stopped out often, and that's exactly where the first test flattered the most. Most of the drop comes from execution: without costs but with realistic fills, Qullamaggie would be at 1.16. Costs then push him below 1.
A second insight came for free. The first test had found that a trailing stop improves all five strategies. We had built it into the screener as a result. Under realistic calculation the fixed stop is slightly better for all five — Minervini, for example, 1.17 instead of 1.15. So we switched the trailing stop off again.
What's left after costs
| Strategy | Trades | Win rate | Avg winner | Avg loser | Avg per trade | PF | Avg days |
|---|---|---|---|---|---|---|---|
| Zanger | 1,563 | 49.4 % | +6.43% | -4.81% | +0.74% | 1.30 | 6.0 |
| Minervini | 12,086 | 39.4 % | +8.91% | -5.02% | +0.47% | 1.15 | 16.1 |
| Confluence | 11,941 | 48.7 % | +5.94% | -5.05% | +0.30% | 1.11 | 12.7 |
| Kell | 12,029 | 32.5 % | +4.89% | -2.61% | -0.17% | 0.90 | 4.2 |
| Qullamaggie | 11,608 | 30.8 % | +4.59% | -2.32% | -0.19% | 0.88 | 4.7 |
The win rate stands out at first glance. Minervini is right only 39% of the time, and the strategy still makes money because an average winner earns almost twice what a loser costs. Qullamaggie and Kell have a similar ratio, they just win too rarely: three out of ten trades aren't enough.
The median is negative for all five strategies. So the typical individual trade loses slightly, and the few big winners carry the result. That's normal for momentum and psychologically the hardest part: if you change the rules after three or four losses in a row, you miss exactly the trades that make the year.
The crucial question: better than random?
A profit factor above 1 sounds good. But it only says that the strategy made money during the test period, and 2023 to 2026 was a strong market. That's why we ran the same test with a strategy that has no intelligence at all: buy every stock in the universe on every day and hold it for 5, 10 or 20 days, with the same costs. Only the distance to that line shows whether a filter is any good.
Better than random? Profit factor after costs vs. holding period
Gray line: buy any stock from the same universe on any day and hold it for 5, 10 or 20 trading days, with the same costs. If a strategy is below the line, it was no better than a random purchase with a similar holding period.
Zanger is clearly above the line. At a holding period of about six days a random purchase gets about 1.0, Zanger 1.30. That's the only clear edge in the test. Minervini and Confluence are roughly on the line: they make money, but no more than arbitrary purchases with a similar holding period. Qullamaggie and Kell are below it.
Zanger's rule set requires a breakout on sharply higher volume — a concrete event on a specific day. The other rule sets describe more of a state: uptrend, proximity to the high, good structure. A state often lasts for weeks and says little about whether today is a good day to enter. We don't think it's a coincidence that the event filter comes out on top.
Why short holding periods lose
Costs explain a large part. 0.2% per buy and per sell adds up to about 0.4% per trade. Before costs, Qullamaggie and Kell only make about 0.2% per trade on average, because they're already out again after four to five days. Costs eat that twice over. Minervini, with a 16-day holding period, makes just under 0.9% per trade gross, and something is left after costs.
On top of that, strong stocks run for weeks, not days. If you enter with a tight stop, normal noise stops you out before the move begins. We cross-checked this: doubling the maximum holding period raises Zanger from 1.30 to 1.43 and Minervini from 1.15 to 1.27. Even with longer holding periods, wider stops or other exit rules, Qullamaggie and Kell get at most just above break-even — and not above the level of random purchases in either half of the test period.
Qullamaggie actually trades mostly smaller, very volatile stocks, often after earnings gaps. So we expanded the universe to about 3,000 US stocks including small caps and tested entries closer to the methods: a breakout from a tight consolidation after a strong run-up, and for Kell the pullback to the 20-day line. That brought no edge over random purchases in the same stocks either. Whatever makes these traders successful apparently lies in things that don't show up in price data: catalysts, themes, picking the best two out of fifty candidates.
Every year is different
| Strategy | 2022* | 2023 | 2024 | 2025 | 2026** |
|---|---|---|---|---|---|
| Zanger | 0.06 | 0.93 | 1.53 | 1.47 | 1.49 |
| Minervini | 0.07 | 1.03 | 1.29 | 1.16 | 1.07 |
| Confluence | 0.05 | 0.94 | 1.32 | 1.01 | 1.15 |
| Kell | 0.13 | 0.83 | 1.01 | 0.79 | 0.97 |
| Qullamaggie | 0.10 | 0.80 | 0.98 | 0.71 | 1.04 |
| * from mid-May 2022, only 15 to 92 trades per strategy, because the market filters screened out almost everything in the bear market. ** through mid-September 2026. | |||||
2022 was a disaster for all strategies, but with very few trades. The market filter screened out almost everything in the bear market, and what got through went wrong. 2024 was the best year, 2025 and 2026 weaker. Zanger is the only strategy that has been clearly above 1 in every year since 2024.
In practice: a single year says little. If you consider a strategy proven after three good months, you've most likely measured a good market, not a good strategy.
The current rules as a portfolio
The screener today runs with two rule changes compared to the first part. The trailing stop is off, because the fixed stop did better in the test. And Minervini and Qullamaggie now require an RS rank of at least 80 for an entry: the stock has to have outperformed 80% of all stocks in the universe over the past year. We set this rule before the test and checked it on two separate periods, 2022 to 2024 and 2025 to 2026.
There is also no fixed maximum holding period here. The screener has no time exit: a position runs until the stop or the target is hit. This part shows what the product actually does.
Profit factors per trade say little about what a trader ends up with in the account. So we also ran each strategy as a model portfolio: $100,000 starting capital, 1% risk per trade, at most eight positions at once. If there are more signals than room, the most liquid stock wins. The benchmark is the S&P 500 over the same period, including dividends.
| Strategy | PF per trade | Random* | Portfolio total | 2022–24 | 2025–26 | Max drawdown | Top-10 share** |
|---|---|---|---|---|---|---|---|
| Zanger | 1.62 | 1.47 | +98% | +59% | +38% | -26 % | 61 % |
| Minervini (RS ≥ 80) | 1.58 | 1.71 | +75% | +70% | -2%*** | -24 % | 70 % |
| Qullamaggie (RS ≥ 80) | 1.15 | 1.22 | +46% | +45% | +10% | -33 % | 127 % |
| Kell | 0.93 | 0.95 | +42% | +3% | +45% | -34 % | 92 % |
| Confluence | 1.36 | 1.76 | +42% | +47% | -1%*** | -25 % | 108 % |
| S&P 500 | – | – | +101% | +52% | +32% | -19 % | – |
| * Profit factor of a random purchase with the same holding period. ** Share of the portfolio profit from the ten best trades; above 100% means the portfolio would be negative without them. *** fewer than 100 trades, not reliable. Portfolio: $100,000, 1% risk per trade, at most 8 positions, no time exit, costs as above. Each period starts with its own portfolio. Bold: better than the S&P 500 over the same period. | |||||||
Only Zanger beats the S&P 500 in both periods, 2022 to 2024 and 2025 to 2026, and it is also clearly above a random purchase with the same holding period on a per-trade basis. The lead is narrow, though — over the whole period Zanger is even slightly behind the index. That comes from the math: the overall figure runs a single portfolio, while each sub-period starts fresh with $100,000.
Minervini lets winners run and was well ahead of the index through the end of 2024. Since 2025 the portfolio is slightly negative, but with fewer than 100 trades — too few for a reliable verdict. Per trade, the rule is below a random purchase with the same holding period: if you hold for eleven weeks, you'd also have done well with randomly chosen stocks during the test period.
Qullamaggie and Confluence trail the index in both periods, Kell is only ahead in 2025/26. For Qullamaggie and Confluence, the portfolio would be in the red without the ten best trades.
A few trades make the result. Even for Zanger, about 60% of the profit comes from the ten best of 242 trades. That's typical for momentum and at the same time the biggest uncertainty: a few missing home runs and the picture looks completely different.
Our honest summary: Zanger has the most robust numbers. All other rule sets are useful as a pre-filter, but as a mechanical buy rule they're no better than an index fund. Over the whole period, none of the five strategies beat the S&P 500.
What this means for you as a trader
A screener hit is not a buy signal. We've always said that, and this test provides the numbers: if you mechanically buy every hit, four out of five rule sets land you at roughly random. The filter saves you the work of going through 900 stocks. It doesn't make the decision about which candidate is really worth it.
We'd take three things from the test. First: always factor in costs, and do it before you get excited about a strategy with short holding periods. Second: give good trades time. A stop belongs where the thesis is proven wrong, not where it feels good; more on that in where to place your stop loss. And third: distrust any backtest that doesn't say how it handles gaps and costs. Our own first one didn't.
Limits of this study
The test has weaknesses, and we'd rather name them ourselves. It uses today's stock list for past years as well, including in the model portfolio. Companies that dropped out of the index or went bankrupt along the way are missing, and that tends to flatter all the numbers. The rule sets are mechanical approximations of the traders' methods, not copies. The period is mostly a bull market. Taxes aren't included, and the fixed maximum holding period in the first part is a simplification the live screener doesn't have.
How we set up data, rules and tests is described in detail on the page How FyndStock works.
The screener finds the candidates, the decision is yours
FyndStock checks more than 900 US stocks four times a day against all five rule sets, including sector and market filters. The shortlist takes seconds, leaving your time for the actual analysis.
Start freeBacktest FAQ
What is a profit factor?+
The profit factor is the sum of all gains divided by the sum of all losses. A value of 1.30 means: for every dollar lost, there were $1.30 of gains. Below 1, a strategy loses money overall. It says nothing about how much results fluctuate, so it should always be read together with win rate, average gain and a benchmark.
Why are so many backtests too optimistic?+
The most common reasons: stops are booked at the stop price even though the stock gapped below it. Costs and spreads are missing. The stock list only contains today’s survivors. And the rules were tweaked until the past looked good. Our first test had the first two flaws and overstated the profit factor by 0.5 to 1.7 points as a result.
Do momentum strategies work at all?+
Momentum is one of the best-documented effects in the stock market. But our test shows that a simple filter on its own often does no better than a random purchase after costs. The edge comes rather from a concrete trigger (here: a breakout on volume), longer holding periods and picking the best candidates — that is, from what a trader does after screening.
Why does Qullamaggie do so badly even though he is successful?+
Because we didn’t test him, but a mechanical approximation of his publicly described criteria on large US stocks. He prefers smaller, very volatile names, often after news-driven gaps, and picks very few out of many candidates. Short holding periods with tight stops are especially unfavorable for a mechanical implementation after costs.
What trading costs should I expect?+
With US stocks at a low-cost broker, commissions are often zero or close to it, but there is still the spread and slippage on execution. We assumed 0.2% per buy and per sell, about 0.4% per trade in total. If you make many short trades, check that number against your own broker.
Can I verify the results?+
The method is described on the page “How FyndStock works”, and each strategy’s rules on the strategy pages. The test uses the same code as the live screener and only sees, on each simulated day, the data that existed at the time.
Keep reading
- Minervini, Kell and Qullamaggie rebuilt: what the original rules achieve in a backtest
- How FyndStock works: data, rules and backtests
- The Zanger strategy: breakouts with volume confirmation
- Where to place your stop loss: distance, ATR and the most common mistakes
- Spotting failed breakouts: why breakouts fail
- Risk/reward calculator: is the trade worth it after costs?
This article is for educational and informational purposes only and is not investment advice or a recommendation to buy or sell. Trading securities involves substantial risk, including the loss of your entire investment. Backtests and past results are not a reliable indicator of future performance.