Nine thousand strategies from three copy trading platforms: how many of them are alive, where subscribers actually put their money, and why figures from different shop windows cannot be placed side by side.
Before choosing a platform for copying trades, I decided to look at what actually goes on there — not at the shop window with the best performers, but at the whole thing. I collected the full rankings of three platforms: Tickmill Social — 1,438 strategies, RoboForex — 5,056, MQL5 Signals — 2,622. Plus the Myfxbook catalogue, 668 systems.
Accounts without a single trade were removed: they drag any statistic down and are not strategies in any case. That leaves 1,407, 4,497 and 2,620 respectively. Everything below is about those.
All of it is a snapshot of one day. The make-up of these platforms changes: accounts appear, go to zero and disappear. A year from now the figures will be different, which is why the date of the measurement stands next to them.
The first thing I stumbled over: the platforms’ figures are not comparable. Each shows a “drawdown”, and none of them writes down what exactly it counts as one. There is only one way to check — take an account published in two places at once and see what each of them says about it. I have such an account: the prototype of the portfolio is published both on MQL5 and on RoboForex. The trades are the same, down to the last one.
| figure | MQL5 | RoboForex |
|---|---|---|
| return | 39.46% | 39.46% |
| balance | $144.77 | $144.77 |
| drawdown by balance | 12.96% | — |
| drawdown by equity | 2.12% | 2.82% |
| trades | 109 | 218 |
The money matched to the cent. After that the divergence begins. RoboForex does not show a balance drawdown at all, and what it calls simply “drawdown” turned out to be the equity one — 2.82% against 2.12% on MQL5. Same metric, same account, and the numbers differ by a third: the platforms sample equity at different intervals.
And the second line, the one people usually skip: 109 trades against 218, exactly double. One platform counts positions, the other counts deals — entry and exit separately. It is not an error but different MetaTrader terminology, and in a ranking it turns identical activity into a twofold difference.
That way I could cross-check two platforms out of three. There is nothing to check Tickmill against: its methodology is not published, and my account was not there yet at the time of the measurement. So its drawdown below stands apart and is not placed in the common row — that is more honest than putting it there and pretending the numbers are comparable.
The most unexpected finding was not in the drawdowns but in the money. The median account in the Tickmill ranking holds four dollars.
| figure | Tickmill | RoboForex | MQL5 |
|---|---|---|---|
| strategies with trades | 1 407 | 4 497 | 2 620 |
| median account | $4 | $327 | $1 846 |
| accounts under $10 | 52% | 4.8% | 0.5% |
| median age, days | 209 | 155 | 154 |
It means half the shop window is abandoned accounts. Someone opened one, lost the money and left; the card stayed and keeps taking part in the ranking. On RoboForex such accounts are 4.8%, on MQL5 half a percent.
The first explanation that comes to mind: the other platforms simply tidy up after themselves and remove accounts with no activity. I checked: they do not. The oldest account in the Tickmill list is 3,607 days old, on RoboForex 3,194, on MQL5 2,877 — nearly ten years, and all of them are still there. There is no money threshold either: MQL5 lists signals with one cent on the account.
The difference is not in the cleaning but in the composition. Among accounts older than two years, 42% are empty on Tickmill, 1% on RoboForex and none at all on MQL5. Nobody sweeps the shop window, but different people come to each of them and stay.
The obvious next thought is that it comes down to admission rules. But the rules are similar everywhere: demo accounts are banned on all three, and Tickmill on top of that has no cent accounts as a class — which means a median of four dollars is not a small denomination but an ordinary account taken to zero. The entry threshold there is higher too: $250 against $100 on RoboForex.
The result is the opposite of what one would expect. The barrier at Tickmill is stricter, and there are three times fewer strategies because of it — 1,438 against 5,056. Yet the share of dead ones among them is ten times higher. The threshold cut the quantity, not the composition: someone who comes to open a $250 account and lose it passes any formal filter. So a shop window is shaped not by admission rules but by who comes to the platform at all, and why.
Figures are figures, but there is a simpler question: how much of his own money does the trader have, whose trades you are invited to copy. The platforms do not hide it — the provider’s account is visible on his card.
| provider accounts | Tickmill | RoboForex | MQL5 |
|---|---|---|---|
| median account | $4 | $327 | $1 846 |
| accounts under $10 | 52% | 5% | 0% |
| accounts under $100 | 57% | 14% | 3% |
| accounts over $1,000 | 20% | 28% | 66% |
| accounts over $10,000 | 3% | 4% | 22% |
The last line is worth pausing on. The share of genuinely large accounts is almost the same on both broker platforms — 3.0% on Tickmill against 3.7% on RoboForex. Serious providers exist on both, in similar proportion. The whole difference between the shop windows sits in the lower part of the list: on RoboForex it is filled with working, if small, accounts; on Tickmill with empty ones. The four-dollar median comes about not because serious people are few, but because they drown in the debris.
MQL5 stands apart in this row: large accounts there are 22%, six times more than on either broker. Its lists are built differently, and in the money of the participants that shows most sharply of all.
And separately — those who have already been trusted with money, that is, strategies with five or more subscribers. On Tickmill their median account holds 93 dollars. Not the abandoned ones, not random ones — the ones chosen by living people. On RoboForex it is $404, on MQL5 $1,218.
Hence a simple check that requires nothing but a look at the card: if the trader has less on his account than the investor intends to put in, he is risking less than the investor. His own drawdown in money will be smaller, and he will sit through it more calmly.
The entry thresholds are arranged differently. On RoboForex the minimum deposit for a subscriber is set by the author himself: 63% put $100, and one occasionally meets a prohibitive hundred million — a polite way of letting nobody in. On Tickmill the minimum is set by the platform, $250, and the author may raise it for his own strategy. On MQL5 there is no threshold at all.
On the other hand, MQL5 has no free signals either: not one out of 2,618, median $30 a month. And this difference matters more than it seems. On both broker platforms the author receives a share of the profit, and only of the gain above the subscriber’s previous account high: if the subscriber is at a loss, the author gets nothing until that loss is worked off. On MQL5 the fee is fixed — thirty dollars arrive in a profitable month and in a losing one alike.
So in the way the reward is built, the broker schemes sit closer to the subscriber’s interest than the subscription one. A caveat: RoboForex also offers a second option — a fee per trade instead of a share of profit; there the incentive shifts again, towards trading more often. What matters is not the size of the rate but what exactly is being paid for.
Curiously, for all the difference between the shop windows, the price of the service on both broker platforms turned out to be the same. The median share of profit is 25% in both places, the mean 24.1% on RoboForex against 24.7% on Tickmill, and even the distribution matches: the largest group is 30–35%, then 20–25%. The rate was set by the market, not by the platform. On RoboForex 94% of offers are a share of profit, the remaining 6% a fee per trade.
I deliberately leave means and maxima out of this table. Parsing the MQL5 pages I saved the account amount but not the currency — and accounts there come in won and dong, where the nominal is a thousand times larger. The median is immune to that: 1,846 against 1,796 with the top percent removed. Means and records, though, such data turns into nonsense. One more case where figures from a shop window cannot be taken at face value.
Let us take four conditions — the ones an investor would put to a strategy before handing over money. Drawdown no deeper than a quarter of the account. Age over a year. At least five subscribers. Return over the whole period above 50%. And apply them to all three platforms at once.
One caveat straight away: these conditions matter more the less verifiable calculation a strategy has behind it. If there is a history you can recompute yourself, a short live record is not so frightening — you can at least see what the system did in the awkward years. If there is nothing but the account itself, it is all you can judge by, and each of the four filters works at full force.
| filter | Tickmill | RoboForex | MQL5 |
|---|---|---|---|
| strategies with trades | 1 407 | 4 497 | 2 620 |
| drawdown under 25% | 212 | 2 002 | 1 930 |
| + older than a year | 27 | 376 | 359 |
| + at least five subscribers | 4 | 129 | 6 |
| + return above 50% | 2 | 24 | 6 |
| share of the living | 0.14% | 0.53% | 0.23% |
Nowhere does it add up to even one percent. That is the main result of this study: the point is not a particular shop window, but that strategies with time, moderate risk and other people’s trust behind them are units per thousand.
A caveat: the first filter is not entirely fair across platforms — each counts drawdown its own way, as shown above. The other three conditions are directly comparable, and it is they that cut the most: on MQL5, out of 359 strategies older than a year, five subscribers were found for only six.
The conditions can be moved and the numbers will change. What matters is the order of magnitude, and it holds: however much you soften them, the survivors remain fractions of a percent.
A reader has every right to ask: if half of this ranking is the residue of zeroed accounts, why did I go there. I will answer in the order in which it actually happened.
First, a technical reason. Half the portfolio is indices: DE40, US500, USTEC, JP225. In my test runs the index history from RoboForex arrived with noticeable gaps in the quote feed, and a five-year calculation cannot be done on a history full of holes: the gaps fall exactly on the moments the calculation exists for. On Tickmill the history for these instruments turned out to be continuous — which is why the whole calculation was done on their data.
After that it became a decision by itself. Five years of history are computed on Tickmill quotes, spreads, commissions and instrument specifications; the settings files are named after the broker accordingly. It makes sense to open the live account in the same place: otherwise the main comparison — live against calculated — would be impossible, and recomputing everything for another broker was not something I was going to do. I went there to verify the calculation, not to find subscribers.
I see the price of that decision: demand on Tickmill is half that of RoboForex — 3.9 subscriptions per strategy against 7.7 — and the neighbourhood is as described above. But as the funnel showed, the neighbourhood is much the same everywhere, while comparability of calculation with a live account exists in one place only.
The second question all this was counted for: do investors choose traders well?
| figure | Tickmill | RoboForex | MQL5 |
|---|---|---|---|
| subscriptions per strategy | 3.9 | 7.7 | 0.2 |
| strategies with 5+ subscribers | 225 | 1 043 | 20 |
| their median return | −52% | +22% | +572% |
On Tickmill, among strategies that gathered five or more subscribers, the median return over the whole period is minus 52 percent. Not the worst ones — the middle one among those trusted with money. A third of all subscriptions on the platform went to accounts with a drawdown deeper than 90%, that is, to those who have already been to zero once.
On RoboForex the picture is milder, on MQL5 the reverse. This is not about platforms deceiving anyone: a ranking simply sorts by return, and return is easiest to obtain by increasing risk — for a while. The lower the bar for getting into the list, the louder those who bet everything and won over a short distance sound in it.
The fourth source was the Myfxbook catalogue, 668 systems. The figures there are pretty: one loss-making system in the whole catalogue, no drawdown deeper than 49.7%. But that is not a market, it is a filter: by default the catalogue shows only systems available for subscription and passed for listing.
Worth remembering every time you see statistics about “the average strategy”. If the sample is selected, the average will be whatever was selected. The full rankings this study starts from are valuable precisely because everyone got into them — including the four-dollar accounts.
Several checks follow from all this. They are simple, and can be applied to any trader and any copy trading strategy — mine included.
None of these checks promises a profit. They cut out something else — the cases where a decision is made blind. Judging by the figures above, blind is how it is made most of the time: the median return of strategies investors put money into is negative on one of the platforms.
I put the same four questions to my own portfolio. The answers are not here but where they belong: the five-year calculation and every trade as a download are on the Backtests page, the live account and independent verification on Results.
The data was collected from the platforms’ open interfaces — the same ones that feed their own lists and widgets. Nothing closed was used and no personal data was gathered: only public strategy figures. The calculation is my own, and its results are this page.