Copy Trading Commission: Does a Higher Fee Buy a Better Trader?

Every explainer on copy trading commission tells you the same two things: the trader takes a cut of your profit, and the cut is a percentage. Then comes a worked example with round numbers, and the article ends. What none of them say is whether the percentage means anything - whether the trader asking 40% is, on any measure at all, better than the one asking 10%.

That question has an answer, and it needs a census rather than an opinion. Our system reads every strategy RoboForex lists on its copy-trading service once a day and keeps the record after a strategy leaves. On 10 October 2026 that was 4,991 listed strategies carrying 33,338 follower seats, each with a published fee, an all-time return, a worst single step and a registration date, and seven in ten also carrying an operator id that ties the account to the person running it. That is enough to ask what the fee predicts, and to answer it the same way twice on two different platforms.

The short version: the fee predicts no better record, a higher chance of the strategy disappearing, and something about the crowd.

What you are actually agreeing to

Before the numbers, the mechanism, because three parts of it do not appear on the strategy card.

It is a performance fee on a schedule, not a haircut on your result. The platform's own documentation puts the range at "5% to 50% of net profit", paid "only if the overall result is positive", and a separate "No Commission" scheme sits beside it. That matches the list exactly: of the 4,991 strategies listed, 4,649 run the performance fee - every one of them at a multiple of five between 5 and 50, holding 31,643 of the 33,338 seats - and 339 run the no-commission scheme. A third scheme exists, a flat subscription charge in dollars per period rather than a share of profit, and three strategies use it.

It settles on a clock, and the trader sets it. The period is chosen from three options - one, two or four weeks - and 87.8% of listed strategies settle weekly, 3.6% every two weeks and 8.6% every four. The commission is reserved on the investor's account rather than transferred as it accrues, and moves to the trader at the end of each period - or, on unsubscribing, usually within a few minutes.

There is a high-water mark, and it is the most valuable line in the offer. The documentation is explicit: "Only profit above the previous highest result (High watermark) is used to calculate your commission", with a worked example - profit of $100 earns the trader $10 at 10%, a rise to $150 earns $5 on the $50 of new profit, and a fall back to $140 earns nothing. Without that clause the same fee costs far more; further down there is a measurement of how much.

One part is less comfortable. Subscription conditions "can be updated once every 24 hours", the change takes effect on Monday at midnight server time, and when a trader changes the rate, in either direction, "the new rate applies to all copied trades (old and new)". The fee you agreed to is the fee as at last Monday. It is not a hypothetical: 2,244 listed strategies - 45.0% of the list - are running an offer that took effect more than a week after the account was registered, and 437 of them moved onto that offer within the last thirty days.

What the list actually charges

The median strategy charges 25%. The average copied seat is signed up at 27.5%, because the seats do not sit where the strategies sit. 30% is the single most common rate, on 1,136 strategies holding 9,099 seats. At the top of the range, 343 strategies charge the maximum 50% - 6.9% of the list - and they hold 12.4% of all follower seats. Taken together, strategies charging 30% or more are 37.4% of the list and carry 55.1% of every copied seat on it. The platform's own count of recently added subscriptions - 3,236 seats over an interval it does not define - points the same way: 62.6% sit on a strategy charging 30% or more.

So most of the copying is happening on the dearer third of the list. The obvious explanation is that the dearer third is better. It is not.

A higher fee does not come with a better record

Rank the 4,991 listed strategies by fee and by all-time return and the correlation between the two is 0.016 - no directional relationship, on a sample large enough that anything much above 0.03 would have stood out. Restrict it to the 2,385 strategies someone is actually copying and it is 0.037, still short of the line that separates a signal from noise. Restrict it to the 2,602 with at least 180 days of record and it is -0.009. Compare the fee with the worst single step the platform reports - which is not the peak-to-trough fall, and flatters every strategy on the list - and it is 0.009.

For a sense of scale, the same calculation on the same list finds 0.224 between follower count and return - plausibly because followers pile onto strategies that have already done well, though a snapshot cannot say which way it runs. That is a weak relationship, but it is a relationship. The fee is not one.

That does not mean the tiers are identical. They are not: a test of whether the nine fee groups differ at all says they do, clearly. What it does not find is a direction. The share of strategies underwater runs 42.5% among the free ones, 34.4% at 20-24%, 39.5% at 30-34%, 49.5% at 40-49% and 42.3% at the full 50%. The median all-time return is +9.8% at 20-24%, +12.2% at 35-39%, exactly zero at 40-49% and +2.0% at 50%. Up, down, up, down - the groups differ, but not in an order, and the dearest end is not where the better records are.

The natural objection is that something else - age, or a crowd that only pays up for proven names - is doing the work, so the same question was asked inside each age band and each follower band. In every follower group from zero to over a hundred, the correlation between fee and return sits between -0.040 and +0.045. Five of the six age groups are just as flat. The exception is worth stating rather than hiding: among strategies 90 to 180 days old, the dearer ones have done better - a median +11.3% for those charging 30% or more against +4.6% for those under 25%, a correlation of 0.108 that holds up even after allowing for having looked in six places. One band of six, in a single month's reading, and the bands on either side of it show nothing. It is a reason to check again on the next reading, not a reason to pay more.

The expensive end disappears faster

The list flatters the expensive end to begin with, because the strategies that leave it are disproportionately expensive. Our panel has seen 8,420 strategies since it started on 25 August; 3,457 of them have since left the list, and 45.8% of those charged 30% or more, against 37.4% of the strategies still listed. Two thirds of them - 66.5% - were underwater the last time we saw them.

Split by fee, the gap is plain. Of every strategy the panel has ever seen at a given rate, the share still listed today:

Two in three free strategies seen since late August are still there; at 35% and above it is about one in two. A follower choosing from today's list is choosing from the half of the expensive end that is left, which is part of why the snapshot above looks as level as it does.

The mirror image is worth stating on its own: 743 listed strategies charge 30% or more and are underwater. That is 39.8% of every expensive strategy on the platform, and 4,886 follower seats are sitting on them.

What the fee does predict

Two things, and the second is weaker than it was.

It predicts followers. The correlation between the fee and the follower count is +0.072, small but far too consistent to be chance. Which way that runs, a single day's snapshot cannot say: the same list has 2,244 strategies on an offer written after the account was opened, so a trader raising the rate once the followers arrived would leave the same trace as followers choosing the dearer trader.

It leans higher on second attempts. Of the 1,042 strategies registered since 1 August that carry an operator id, 293 belong to an operator whose earlier account had already lost half its value or taken a single fall of 90% or more - as far back as our record goes, 25 August, so a floor rather than a count. That is one in five of every strategy listed since 1 August, and nearly three in ten of those we can trace to an operator at all. Their median fee is the same 25% as the other new listings, but more of them sit at the expensive end: 46.8% charge 30% or more, against 40.6%. Between them they hold 2,447 seats.

One thing it no longer predicts is the trader's own stake. On the first reading of this list the balance on the trader's own account ran a little lower where the fee was higher. It still does across the whole list - a median $326 where the fee is under 25% against $271 at 30% or more - but among the strategies anyone is actually copying the relationship is gone entirely, at -0.002.

Run the question the other way and the point lands harder. Filter the list for the record a careful follower would actually accept - in profit all-time, a worst single step better than -15%, and at least 180 days old - and 318 strategies survive, 6.4% of the platform. Their median fee is 25%. That is the same as the list as a whole, and the share of them charging 30% or more, 38.4%, is the list's own 37.4% give or take a point. Quality, where it exists, is not priced.

The rate on the card is the floor, not the price

A percentage taken again and again is not a percentage taken once: every settlement removes money that would otherwise have kept compounding, so the trader's share of your final gain grows with the size of the gain and the length of the record, as the copy trading fee calculator shows for any stake and horizon - before anything else happens at all.

We took the 100 most-followed strategies, pulled each one's full published yield curve, and ran the fee through it under the platform's own rule: settle on that strategy's own period, charge only on cumulative profit above its previous high, and settle once more on the way out, as the platform does when a follower unsubscribes. All 100 had a usable curve, between them carrying 16,981 seats, and 80 of those ended their curve in profit - the only ones for which "a share of the gain" means anything. The model has the follower joining on the curve's first day and holding to the last: the longest exposure to the fee, and the only entry that can be read off a public curve without inventing one.

Across those 80, against a median headline rate of 30%, the median share of the follower's entire gain taken by the fee was 45.7%. Weighted by seats it was 49.1%. 78 of the 80 cost more than the number on their card, and in two of them the fee took more than the whole of the gain.

Most of that gap is arithmetic, not drama. Replace every curve with a perfectly smooth line of the same total return over the same dates - no dips, nothing given back and re-earned - and re-run the same rule: the median share taken is still 38.0%, and 78 of the 80 still cost more than their headline. The compounding alone does it. Vendo DD50 is the clean case: 30% on the card, 50.7% of the gain taken, and 48.3% on a smooth line of the same return. S1 AUDCAD/EURGBP is the same story lower down the scale: 15% on the card, 19.1% taken, 19.0% smoothed.

Giving money back and making it again is the second mechanism, and it is what separates the ordinary cases from the ugly ones. Its median contribution across the 80 is about a point, but 20 of them pay more than five extra points for it, and at the top of that group it dwarfs the arithmetic. FFX STradegy would have paid 27.2% of its gain on a smooth line; on its real curve it paid 76.4%. Albert Einstein would have paid 38.0%; it paid 60.5%. Those are strategies that made money, gave it back, and made it again, each new high paying the trader once more - the shape a martingale-style strategy produces on purpose.

What that looks like from the follower's side:

• 10K To 1M JOURNEY (1,926 seats, 35%, 49 days of curve): +94.0% for the strategy, +52.6% for the modelled follower.
• FFX STradegy (422 seats, 25%, 1,322 days): +25.7% became +6.1%.
• Vendo DD50 (353 seats, 30%, 845 days): +427.5% became +211.0%.
• S1 AUDCAD/EURGBP (254 seats, 15%, 1,207 days; not in the chart above, which shows only the seven most-followed): +79.8% became +64.5%, the smallest gap of the four.

That is also what the high-water mark is worth. Run the same strategies without it - a fee on every profitable period, losing periods forgotten - and the median share of the gain taken rises from 45.7% to 66.9%, with the cases where the fee exceeds the entire gain going from two in 80 to 14. Under that rule FFX STradegy's modelled follower finishes at -16.4% on a strategy that made 25.7%.

And the watermark does not make the fee free of risk. 18 of the 100, carrying 2,849 seats, ended their curve below where they started and had still paid the trader along the way, because they made new highs first. The clearest is the second most-followed strategy on the list. When this article was first published, on the reading of 12 September, HONEY MAKER (804 seats, 50%) stood at +15.2% for the strategy and +4.4% for the modelled follower. Four weeks later the strategy has given back its whole gain, to -0.4%, and the follower who held from the start is at -9.8%, having paid fees worth 10.5% of the original stake on the way up. On TRDR XAUUSD BOT (325 seats, 30%) a strategy down 8.7% leaves the modelled follower down 13.0%.

Three things this model is not. The curve is the trader's own account, not any follower's statement - proportional copying tracks it closely but not exactly, and nothing here charges spread, swap or platform costs on top. The curves are close to daily without being complete, so some settlements fall inside gaps of several days. And the exit settlement matters less than it sounds: only 17 of the 100 curves pay anything on the way out, and it moves the median share taken by 0.4 of a point.

The same question on a crypto platform

Bybit's copy trading prices itself completely differently, and gives much the same answer.

Reading every Bybit leader with at least one follower on the same day - 1,797 of the 7,041 on the board, holding 17,320 seats - the median profit share is 10%, and 87% of them charge exactly that. The clustering is structural: the ratio is capped by the trader's rank, and the rank table on Bybit's profit-sharing page puts Cadet and Bronze at 10%, Silver at 12% and Gold at 15%. Reality is untidier than the documentation at both ends - 148 leaders charge less than 10% and 83 of those charge nothing at all, while 62 charge more than the 15% ceiling, some of them 50% - and we have no explanation for either group that we would stand behind.

The average copied seat there is signed up at 10.17%, against 27.5% on the forex one. Those are not the same price, though. Bybit's Classic copy trading has no high-water mark: losses offset profits only inside a single Saturday-to-Friday cycle, and a losing cycle refunds what was pre-deducted rather than carrying anything forward. That is the third bar in the chart above: on the forex curves it was the difference between 45.7% and 66.9% of the gain. A lower rate on a card, under a rule that starts from zero every week, is not straightforwardly cheaper.

It buys next to nothing. The correlation between a Bybit leader's profit share and their 90-day return is 0.055 - tiny, a third of one percent of the variation, and only borderline: it clears the usual bar on its own, but not once you allow for the seven relationships we checked on this board. Against the platform's own published figure for what their followers earned over those 90 days - a number the forex platform does not publish at all - it is 0.038, which is not. With seven leaders in eight on the same rate there is little variation to correlate against, so this is a thinner test than the forex one. The relationship that does show up is the familiar one: profit share against follower count, +0.069.

Split into three groups, the picture is descriptive rather than a tested result, and the groups are very uneven in size. The 85 leaders charging above 10% show a better median 90-day return of their own, +5.2% against +1.6% at the standard rate, and a shallower median drawdown, 17.2% against 25.7%. Their followers' median P&L is -$21, with 55.3% of them negative. The 148 charging below 10% show -$8, with 56.1% negative; the 1,564 at the standard rate show $0, with 41.1% negative. So both ends look worse for followers than the middle, and at the dearer end the trader's own curve and the followers' money point in opposite directions - which is the whole problem with reading a leader's return as a promise. Across all 1,797 leaders those P&L figures sum to about -$1.31 million over the period, with 773 of them in the red - a loss we have already totalled leader by leader. Here it is enough that the cheaper platform is not the profitable one either.

What to do with this

The fee is the one term of a copy-trading subscription that is certain. The return is not. Everything above says the two are unrelated, so the fee should be treated as a cost to minimise rather than a signal to read.

Three practical consequences:

• Read the settlement period, not just the rate. A 25% fee settling every 28 days and a 25% fee settling weekly are different prices for the same strategy, and the difference grows with how much the strategy moves.
• Price the fee against the curve, not the card. What it takes grows with the size of the return, the length of the record and every dip made back: above, a 30% card cost 48% of the gain on a smooth curve and a 25% card cost 76% on a jumpy one.
• Check whether the rate you signed up for is still the rate. It can be changed once a day, takes effect on Monday, and applies to trades already open.

None of which decides whether to copy anyone at all. That still comes down to the things the whole-list statistics already showed - a third of the list underwater, a median barely above zero, survivorship doing the flattering - plus how much of your money one strategy gets. The fee is just the part you can work out in advance, and it is larger than it looks.

Every figure above is our own computation across the figures RoboForex and Bybit publish per account, as recorded on 10 October 2026; the article was first published on 12 September with that day's reading. The RoboForex panel has read the full list daily since 25 August 2026, and the Bybit universe is every leader with at least one follower on that day's board. RoboForex returns are all-time, Bybit's are 90-day, and the two are not directly comparable. The fee simulation is a model of the published rule against published curves, not any real account statement. Shares and counts will move with every reading. Copy trading carries real risk, and nothing above predicts future performance.