Challenge Distillation: Run Multiple Accounts on Unrelated Waves, and Only Advance the Survivors | Same Expected Value, but the Odds of Actually Winning Went from 13% to 81%

※ This article is opinion/commentary. It is not investment advice. Simulations are model estimates and do not guarantee future results. Firm-specific rules change frequently — always check the official site of each firm. (Verified: September 17, 2026)

Until May 2026, I mirrored the same trades across every account. With 10 accounts, all 10 would move in the same direction at the same time. Now I do something completely different.

What is challenge distillation

Let me give you the shape of it up front.

1. Buy multiple challenges at the same time and assign a different strategy to each account (don't copy the same trades)
2. Only advance the accounts that survive
3. Repeat the same process within that surviving group, again with separate strategies

0000000000
00000
00
0

That's it. I call this "challenge distillation."

And I run it like a campfire. If an account blows up, I refill it immediately and keep the cycle turning. Profits get plowed back into more challenges and more wave diversification. Keeping the fire burning is the operation itself.

At first I suspected I was just manufacturing survivorship bias. Pointing at the survivors and calling them "strong" is one of the most classic statistical mistakes there is.

So I checked it with 40,000 Monte Carlo runs. The short answer: half of my suspicion was right, and half was wrong.

What actually changed between May and now

The number of accounts didn't change. What changed is the number of independent bets.

Mirroring every account means that even with 10 accounts, you're making 1 bet. Since you're copying the same trades, the correlation is 1. All 10 accounts win together and lose together. This is mathematically identical to "stacking 10x the chips on a single bet" — it isn't diversification. It's concentration.

Distillation means 10 accounts run 10 separate strategies, so you're making 10 bets. Correlation is near zero. If one dies, another can still be alive.

Same "10 accounts," but one is a single bet and the other is 10 separate bets. That's the only difference. And that difference dominates the outcome.

With 20 accounts, distillation's relative risk drops to 22% of a single account. Mirroring stays at 100% no matter how many you stack.

What matters isn't the count — it's "how spread out the waves are"

This is the crux of it. Increasing the number of accounts does almost nothing if the correlation between them stays high.

The relative risk of N accounts (with a single account = 1.0) is determined by the correlation ρ between accounts as follows:

Relative risk = √( (1 + (N−1)ρ) / N )

The lower bound is √ρ. Stack hundreds of accounts at correlation 0.8, and risk never drops below 0.894. At 20 accounts it's 0.900. Basically meaningless.

By contrast, at correlation 0 it drops to 0.224 at 20 accounts. What matters is the correlation, not the count. That's why "a genuinely different wave on every account" is the lifeline here.

This isn't just a theoretical formula. I built 40 actual strategies from real price data — 5 strategies × 8 instruments — and measured them. The average correlation between strategies came out to 0.007 (effectively zero). Individual pairs ranged from as high as 0.88 (running the same strategy on the S&P 500 and the Dow looks similar, of course) down to as low as −0.74 (trend-following vs. mean-reversion on the same instrument), and it nearly cancels out on average. The full write-up is in Is Diversification Really a Free Lunch?.

There are plenty of ways to split waves

"I don't have 10 separate strategies" — I think this is the first wall people hit. But what you need isn't 10 strategies, it's 10 different waves.

Ranked by how much they lower correlation:

Even a small SL/TP tweak changes the wave. Enter at the same time in the same direction, and one hits TP while the other gets stopped out on the same wick. Once outcomes diverge, correlation drops below 1.

But the effectiveness varies enormously by method. If you only tweak SL/TP and stay at correlation 0.8, risk is still 0.900 at 20 accounts. Before adding more accounts, check whether the waves are actually spread apart. Splitting by day of week is only notable for being "near-zero cost with low correlation" — it isn't the only method.

Warning: don't split by direction

"Account A goes long only, Account B goes short only" looks like a clean diversification, but doing it on the same instrument in the same time window makes it indistinguishable from cross-account hedging, discussed below. Avoid it.

Splitting by day of week doesn't lower your pass rate

You might think "if trade frequency drops to a fifth, it'll be harder to pass." For a plan with no deadline, the pass probability doesn't change.

Whether you hit the target first or the drawdown first is a gambler's ruin problem, and the answer is determined only by the distribution of a single bet and the two boundaries. The pace of betting doesn't affect the outcome. Something that takes 3 months trading 5 days a week takes 15 months trading 1 day a week. Only the time changes.

Note that some firms have a rule that an account expires after 30 days without trading. Splitting by day of week is fine since you'll still trade about 4 times a month, but check the rules if you split more finely than that.

Prerequisite: it's pointless if you're not winning per account

Before talking about distillation, there's something to confirm first: is the expected value per account positive? Stack 10 copies of a negative-expected-value account and you just lose 10x faster.

I ran a 2-phase $100K challenge 40,000 times at 50% win rate, RR 1.0 — completely zero edge. 3 trades a day, 1% risk per trade, 120 trading days.

Every plan's multiplier came out above 1.0, even at zero edge.

The reason is that a challenge is a call option. You lose only the fee if you fail, but on a win the funded-account upside is uncapped. This asymmetry leaves a positive expected value even for a "50% win rate, RR 1.0" trader with no edge at all.

The multiplier being above 1.0 is the sole precondition for distillation. If that breaks down, everything after this is moot.

The main event: what is distillation actually producing

Same conditions, only the number of accounts changes. Plan: Fintokei Sapphire ($529).

Expected payout per account was perfectly linear. $942 / $963 / $973 / $966 per account — the same within margin of error. Distillation doesn't generate a single extra yen of expected value. Just as I suspected.

What was different is the rightmost column. The probability of ending net-positive jumped from 13.0% to 81.1%.

This is the decisive part. At zero edge with 1 account, the expected value is +$942, but only 13.0% of the time do you actually end up net-positive. Nearly 90% of the time you just lose the $529. It's a lottery ticket with a positive expected value that you almost always lose.

And a lottery-shaped expected value is functionally the same as no expected value at all for someone who can't play it repeatedly. If you only get to draw once, 90% of the time you end up with "a loss that, in theory, was supposed to be a win."

Stack 20 of them, and the expected value per account doesn't change, but you now have a business that ends net-positive 4 times out of 5.

Distillation is not a device for creating edge. It's a device for converting an expected value that already exists into a form you can actually collect.

When you have edge, things get simpler

Plug in a 55% win rate and RR 1.2:

With 5 accounts, not a single run out of 40,000 ended net-negative. Distillation doesn't create edge, but for someone who already has edge, it nearly eliminates the uncertainty. Conversely, for someone with no edge, all it does is take "9.4%" to "76.5%" — you don't get something from nothing.

Verified against real data

Everything above is coin-flip (synthetic random number) testing. Does the same thing happen with real markets? I checked it against real prices from July 2010 to September 2026.

I applied 5 standard strategies (moving-average crossover / Bollinger mean-reversion / Donchian breakout / RSI mean-reversion / momentum) to 8 instruments (gold, Nasdaq, S&P 500, Dow, crude oil, USD/JPY, EUR/USD, silver) to build 40 waves. Of the 40, 25 came out positive and 15 negative — a realistic-looking lineup.

I ran these through a 2-phase challenge, randomly picking real calendar start dates and running it 4,000 times. The real correlation between waves and the real quirks of price action carry through as-is. The full write-up is in Is Diversification Really a Free Lunch?.

Running the same wave on every account doesn't move the odds at all

Not a single digit moved. Run the same wave over the same period and every account moves identically, so paying $10,000 for 20 accounts just means you bought 20 copies of a single account.

The claim "it's not the count, it's how spread out the waves are" showed up exactly the same way in real data.

Splitting into more waves cuts the wipeout rate by a factor of 32

So I fixed the account count at 10 and only varied how many separate waves were used. Fees are always $5,000.

Expected payout stayed roughly flat while the wipeout rate dropped by a factor of 32. Same conclusion as the coin-flip test — distillation doesn't generate expected value, it converts it into a collectible form.

And splitting into just 2 waves alone cuts the wipeout rate by 21 points. The value of splitting the first 1–2 waves is enormous, and the returns diminish from there. There's still value in going up to 10, but start with 2.

A new finding: picking "good-looking" waves breaks the diversification

This one I didn't expect. Selecting recently high-performing waves to run together produces this:

The moment you select by performance, correlation jumps 20x to 70x.

The reason is obvious once you think about it: a wave with strong recent performance is a wave that fit the recent market regime. Things that respond to similar conditions move similarly. "A period when gold was strong" fills the top ranks with gold-related waves; "a trending market" fills it with trend-following systems.

The results showed the same pattern. Compared at 20 accounts:

Random selection won on every metric. The loss from higher correlation outweighed the gain from being individually stronger performers.

When you gather things that "look good," you think you're diversifying but you're actually concentrating. Pick waves based on "how different they are," not "how strong they are."

Two levers drive this. Don't conflate them

Sorting through everything above, the structural effect breaks down into two independent levers. Conflating them makes it impossible to tell what's actually working, so let's separate them.

Expected payout ≒ Budget × (multiplier − 1)   ← multiplier is set by "fee ratio"
Probability of net-positive is set by count   ← count is set by "budget ÷ fee in absolute terms"

• Multiplier (= expected payout ÷ fee) is determined by what percentage of the account size the fee is. Not the absolute amount
• Probability of net-positive is determined by count. This one is set by the absolute amount

These two move independently, and can even point in opposite directions. Let's look at real plans. Budget: ¥1,000,000. Skill held identical across every case (50% win rate, RR 1.0 = zero edge).

Buying one Emerald loses on both levers

Fintokei's Emerald and Sapphire have identical rules. Target 8% → 6%, max drawdown 10%, daily 5%, split 80%. The only differences are account size and price.

Even so, here's how it plays out:

• The fee ratio is worse. Emerald at 0.620% vs. Sapphire's 0.549%. Same rules, but the bigger account costs more per dollar of account size. This directly produces the multiplier gap (2.41x vs. 2.72x)
• The count drops. With ¥1,000,000 you can only buy 3 Emeralds. With Sapphire, 9

You lose on both expected value and certainty. +¥1,305,720 vs. +¥1,674,888; 31.5% vs. 60.9%. "One big account" is the worst way to buy into this approach.

Sometimes the two levers point in opposite directions: Ruby vs. Sapphire

Looking at Ruby, on the other hand, the two levers move in opposite directions.

Ruby has the worst fee ratio at 0.698%, and its multiplier of 2.14x loses to Sapphire. But because each unit is cheap, you can buy 14 of them, and the net-positive rate of 66.2% beats Sapphire's 60.9%.

If you want expected value, take Sapphire. If you want certainty, take Ruby. Which is correct depends on whether you'd rather "win big" or "not lose." The two levers are independent, so looking at only one will lead you astray.

Discounts are the only lever that moves both axes at once

When the fee drops, the fee ratio falls (raising the multiplier), and the absolute price falls (raising the count) at the same time. This is the only move that improves both levers simultaneously. That's why it's worth waiting for a sale.

Skill vs. structure: which matters more

To be clear, skill has a much higher ceiling. With the same Sapphire, going to a 55% win rate and RR 1.2 multiplies the multiplier by tens of times. The effect of edge dwarfs anything structural.

Even so, there are 3 reasons to fix the structure first:

• You can act on it today. Just pick a plan, apply a coupon, and decide the count
• It's guaranteed. "Pick the plan with the lower fee ratio" always works. "Raise your win rate by 5 points" is not guaranteed to be achievable
• It multiplies. Structure multiplies onto whatever result you get per account, so it helps even more if you do have edge

This isn't an argument for skipping skill-building. But with the exact same budget and the exact same skill, expected payout moves from +¥1,100,392 to +¥2,342,448 purely based on which plan you pick. That gap can't be closed by trading skill, and it keeps following you around even after you do improve.

Funded accounts are also subject to distillation

By this point you might read this as "reaching a funded account is the goal." It isn't. Distillation doesn't stop there.

I also withdraw from funded accounts as fast as possible and let them blow up. Because I believe the expected value lives entirely in passing the challenge.

Why "defending the account" is the wrong move

Compare a funded account to your own capital:

• You only keep 80–90% of profit (100% with your own capital)
• Hit the drawdown and you lose it permanently (with your own capital, you can trade again tomorrow even after a loss)
• Subject to daily loss limits, consistency rules, news restrictions, and position-count limits
• You don't own the position, and payment depends on the counterparty's solvency and goodwill

The downside is capped, but the upside is cut too. This is an option, plain and simple. And the value of an option lies in its convexity, and convexity only turns into cash when you exercise it. Hold it to expiry and time just erodes the value.

A funded account you're "defending to the death" is just a degraded version of your own capital account — more restrictions, a smaller cut, and once it dies, it's gone for good.

Three reasons

1. Expected value is concentrated in the first payout.

The value of a funded account is "the sum of all payouts you receive before it dies." This forms a geometric series conditioned on surviving each cycle. The first cycle starts with a full buffer, so it has the highest survival probability, and each cycle after that is conditional on having survived the first. The further out you go, the thinner the probability, so expected value is front-loaded.

2. Some firms wipe out the buffer on the first withdrawal.

E8 Markets' E8 Pro states this explicitly in its official help center. The moment you request a payout, the static drawdown line permanently shifts to the initial balance. And you can only claim 50% of (balance − initial balance). In other words, from the second cycle onward you're structurally worse off than the first cycle — the rules themselves penalize "defending the account."

3. And the real driver: counterparty risk.

This is the biggest one. In 2026 alone, FundedNext banned EAs entirely on accounts $50K and above, ATFunded suspended service, and MyFundedFX (Seacrest) vanished.

The buffer is "a number inside someone else's account"; withdrawn cash is "an asset inside your own wallet." Same face value, completely different things — if the firm collapses, the former goes to zero. Withdrawing quickly isn't about expected value; it's about moving where the asset physically sits.

The numbers make the cycle turn

Our real, received The5ers first payout was $1,230.86. Meanwhile, the Fintokei Sapphire fee is $529.

One payout funds 2.3 more challenges. As long as this loop keeps turning, the motivation to defend an account weakens. That's the fuel for the "campfire." Blow up an account, refill it immediately; take the cash you withdrew and plow it into more challenges and more wave diversification.

An exception: firms where a static drawdown line never moves

There's one caveat. At a firm like Breakout Prop, where the max drawdown is completely static and never moves even after a withdrawal, the picture changes.

On a $100K Classic account, the breach line is permanently $94,000. Grow it to $110K and leave it alone, and your buffer is 16%; withdraw and reset it back to $100K, and the buffer shrinks to 6%. In other words, at this type of firm, not withdrawing is the act of buying yourself a buffer — the opposite of a firm like E8 where the line moves.

Even so, I still prioritize withdrawing, because I weight avoiding counterparty risk more heavily than the value of the buffer. Even a 16% buffer goes to zero if the firm collapses. But this is a choice built on "I don't trust the counterparty," not something that automatically falls out of the expected-value math. Keep that distinction in mind.

What you give up: scaling

There's one more thing this approach abandons: scaling plans.

The5ers Growth, City Traders Imperium's VIP track, Hantec (where passing lets you move up to the next size challenge), and Alpha Capital are all designed so that long-term survival grows your account size over time. A blow-it-up-by-design approach abandons this path from the start.

Running $100K on repeat and growing to $400K are two different strategies. The smaller a firm's aggregate cap, the smaller the cost of giving this up (Breakout caps out at $200K, so there's not much room to grow into anyway), so depending on the firm you pick, this can be effectively free.

We're not traders — we're strategists. The job isn't nurturing one account; it's rotating an asset called "an account" and converting it into cash.

This is not "assigning the loss to an account in advance"

"Aren't you just deciding in advance which account is going to lose and dumping it there?" I get asked this a lot.

I can't. Because there's no way to know which account will win ahead of time. To be clear, I never run opposing positions against myself.

There is exactly one form where "dumping the loss" actually works, though, and that's cross-account hedging. Long gold on Account A, short the same size on Account B. Whichever way the market moves, one of them necessarily wins and the other necessarily loses, so no prediction is required. The loser forfeits the fee, the winner becomes funded. That's what "dumping" actually looks like.

It does work. On a Fintokei Sapphire (Phase 1 target 8%, daily 5%, max drawdown 10%), when A hits +8%, B is at -8%. B blows up first on the 5% daily limit but was a sacrifice from the start. Two fees buy you one guaranteed funded account. Normal purchase odds of reaching funded are 17.3%, so this is 5–6x more efficient. That's exactly why every firm names it explicitly and bans it.

Detection is cluster analysis: simultaneous opposing positions, device ID, IP, millisecond-level timestamp matching, payment method, VPS, duplicate KYC, and cross-firm data shared via third-party vendors. "They can't see it because it's a different company" doesn't hold. Penalties are payouts frozen and profits forfeited at withdrawal time, account closure, and a ban on re-registering.

Distillation and hedging point in opposite directions

The self-check is one line: "Do I know which account will win in advance?" If yes, it's hedging. If no, it's distillation.

⚠️ The risk of looking the same from the outside

Even without intent, this can look identical to monitoring systems. Multiple accounts typically share the same IP, the same VPS, the same KYC, and the same payment method — the raw material for cluster detection is already in place.

Four countermeasures: don't run the same instrument on multiple accounts within the same firm / if two accounts do hold the same instrument, make sure they're facing the same direction / keep any hedge-like logic confined within a single account (hedging within one account isn't banned) / split by day of week. The last one is the cleanest — if the days differ, positions never end up open at the same time to begin with.

If you're actually going to do this

1. Pick plans with a low fee ratio (fee ÷ account size). A multiplier above 1.0 is the sole precondition
2. 2-phase plans with a 4–5% daily band. Below 3% daily, three losses in a row at 1% risk each will mechanically kill the account
3. One big account is the worst move. With the same budget, go with many small accounts instead
4. One wave per account. Never mirror
5. Any way of splitting waves works (instrument, day of week, time of day, entry timing, SL/TP). But correlation is what governs the effect. Changing only lot size is not diversification
6. Diversify across firms. This helps with both the regulatory cap and the outward-appearance risk
7. Once an account reaches funded, withdraw promptly
8. Refill immediately after a blowup. Keep the fire burning
9. Don't read survivors as "good waves"

Where it gets dangerous

The single biggest waste is increasing the count without measuring correlation. At correlation 0.8, even 20 accounts leave risk at 0.900. Thinking "these are spread out" when they aren't is the most dangerous mistake.

The rules themselves have ceilings. Fintokei's 3% rule warnings accumulate across accounts (10 warnings ends your service), and Breakout Prop bans trading multiple accounts from the same IP. There are also aggregate caps, so you cannot pile every account onto a single firm.

The model is optimistic. Minimum trading days, consistency rules, fees, payout cycles, and the eventual death of every funded account are not implemented. What you should trust is "the ranking of multipliers" and "the structure by which diversification collapses" — the absolute figures are illusory.

Summary

• Mirroring and distillation have the same account count but a different number of independent bets (1 vs. N)
• What matters is correlation, not count. The floor on relative risk is √ρ. At ρ=0.8, stacking accounts never gets you below 0.894
• Any way of splitting waves works, but the method matters a lot for correlation. Changing only lot size is not diversification
• Distillation doesn't increase expected value. What it increases is "the probability of actually capturing that expected value" — from 13.0% to 81.1% even at zero edge
• The other lever is the fee ratio. The multiplier sets expected value; the count sets certainty. One big account loses on both
• Funded accounts are also subject to distillation. Withdraw and let them blow up, refill immediately, keep the fire burning
• This points in the opposite direction from cross-account hedging (correlation −1 vs. correlation 0). But because it can look the same from the outside, split by instrument and by firm

The verification script is at scripts/sim-portfolio-2phase-100k.py.

FAQ

Q. Does distillation make me start winning?

No. Expected value is set per account, and adding more accounts only scales it linearly. In the test, the expected payout per account stayed roughly constant at $942 / $963 / $973 / $966. The only thing that increases is "the probability of actually ending net-positive." If the per-account multiplier is below 1.0, stacking more accounts just means losing more.

Q. How many accounts should I run?

Check the correlation before you worry about the count. The floor on relative risk is √ρ. At correlation 0.8, risk stays around 0.9 whether you run 20 accounts or 100. Near correlation 0, it drops to 0.224 at 20 accounts. Adding accounts that aren't actually spread apart is just wasted money. At correlation 0 and zero edge, the probability of ending net-positive was 47.8% at 5 accounts, 65.4% at 10, and 81.1% at 20.

Q. I don't have 10 strategies.

What you need isn't 10 strategies — it's 10 different waves. Split by instrument, day of week, time of day, entry timing, or SL/TP — any of them works. But the effectiveness differs. Splitting by day of week is excellent because correlation stays low and it's basically free. Changing SL/TP tends to leave correlation higher, so the payoff from adding more accounts shrinks. Changing only lot size is not diversification.

Q. Is buying one big account a bad idea?

It's the worst way to buy into this approach. Fintokei's Emerald (¥50M) and Sapphire (¥20M) have completely identical rules, but the fee ratio favors Sapphire (Emerald 0.620% vs. Sapphire 0.549%), and you can buy 3 Emeralds vs. 9 Sapphires with the same budget. You lose on both expected value and certainty (+¥1,305,720 vs. +¥1,674,888; 31.5% vs. 60.9%).

Q. Is this a trick to decide in advance which account loses?

No — it isn't even possible. There's no way to know which account will win ahead of time. The only way "dumping the loss" actually works is cross-account hedging (holding opposite positions on the same instrument across two accounts at the same time), which requires no prediction at all — but every firm names it explicitly and bans it. The self-check is one line: "Do I know in advance which account will win?"

Q. If they're separate waves but happen to end up in opposite directions by chance, is that safe?

Even without hedging intent, monitoring systems judge by shape. If multiple accounts share the same IP, the same VPS, and the same KYC, and hold opposing positions on the same instrument at the same time, that becomes raw material for cluster detection. Breakout even includes correlated assets in its scope. It's safer to split by instrument across firms, or to split by day of week so positions never line up. Note that hedging within a single account is not banned.

Q. Shouldn't I defend a funded account?

I prioritize withdrawing and don't play defense. The reasons are that expected value is concentrated in the first payout, some firms — like E8 Pro — are designed so that the drawdown line permanently shifts to the initial balance on the first withdrawal, and if the firm collapses, the balance sitting in the account goes to zero. In our own track record, one payout funded 2.3 more challenges. That said, at a firm like Breakout, where the drawdown line doesn't move even after withdrawing, leaving the balance in place acts as a buffer, so the calculation changes.

Q. Does a surviving wave mean it's a "good wave"?

No — this is the single most dangerous misreading. Even with 10 accounts at zero edge, there's an 85.0% chance that at least one survives. Survival is not proof of skill. This is the same structure as an "incubator fund" that launches a large number of small funds, closes the losers, and raises money on the track record of the survivors. Distillation is a capital-efficiency device, not a wave-selection device.

Related articles

• Is Diversification Really a Free Lunch? Verified Across 40 Real Strategies
• When Should You Change Your Prop Account Settings
• Multiple Accounts: "Rotation Diversification" Beats "Full Mirroring"
• What Dan Cheung Means by "Rotation"
• Beginner's Guide: What Is Prop Account Rotation
• Consistency Rule Simulation
• Fintokei Complete Guide