Rules-based · Backtested · Run walk-forward, out-of-sample
Crypto allocation by the rules, not by instinct — with the whole test record on the table, losses included.
Two mechanical systems: one follows crypto trends, the other accumulates Bitcoin when a model reads it as statistically cheap — designed to behave differently in the same market. Each publishes one decision a day, identical for every subscriber. Subscribers place their own trades on their own exchange accounts; Allocater never holds funds, never holds keys, and never trades for anyone.
Kratos, six-year walk-forward simulation on data the system had not seen: a hypothetical $10,000 ends at $5.82 million, against $69.5k for the same $10,000 held in Bitcoin. Simulated result on historical data — no money was invested and nobody earned it. Along the way: a −49% drawdown and a losing year, shown rather than hidden.
For investors who will place one trade a day by the rules — not day-trading, not set-and-forget.
9 AUG 26public forward log opens — Athena from 7 Aug 26
Every number above is checkable, and the forward log is the one that will matter most — which is also why it proves the least today: it is the newest thing here. Recording begins with the signal for 9 August 2026 for Kratos and 7 August 2026 for Athena, and we will keep saying how short it is until it is long enough to matter. From those dates each system writes its daily decision and never edits it afterwards; rows publish on this page once they age past the two-week public delay. Rows dated earlier are labelled "replay" — today's engine re-run over past completed days, which is context, not a record. When accounts open, a free account will bring that to 3 days for Kratos and 7 for Athena, and a subscription will remove the lag entirely.
Retail crypto is priced by people with better tools.
The retail deck is stacked. Most "edges" sold in this category are noise, hindsight, or outright fiction.
01
Emotion is the worst position in the book
The plan is written down. Then the candle turns red and the low gets sold, or green and the high gets chased. Discretion leaks returns.
02
No real edge vs. institutions
Funds run systematic, backtested models with teams and infrastructure. The retail alternative is a group chat and a gut feeling.
03
Drowning in scam signal groups
Screenshots of wins, none of the losses. Unverifiable "calls." Paid pumps. This category has earned its reputation, and skepticism is the correct response to it.
Allocater is the opposite of that: mechanical rules that can be inspected, results shown with the losing periods included, and nothing that has to be taken on faith.
Why not just…
The other optionsThe usual catchWith Allocater
Follow a signal groupUnverifiable calls — wins screenshotted, losses quietly deleted.Every decision logged publicly before the outcome is known, every losing period shown, every row labeled for what it is.
Build it yourselfYears of research and thousands of backtests before it is trustworthy.2.3 million+ backtests across 12 generations, run walk-forward on data the system had not seen — already done.
Hand it to a fundLock-ups, management fees, no custody, and the rules are never disclosed.Custody stays with the subscriber, execution stays with the subscriber, and every decision is checkable against the public log.
Just hold BitcoinFull 80%-plus crashes, and no plan for when BTC is wildly expensive.Athena ended about 8× buy-and-hold at a shallower drawdown (−59% vs −83%).
Methodology
How a system decides
No forecasts, no feelings — just measurement and probability. Here is how a system turns raw market data into a position, the same way every single day.
01
Measure, never predict
Each day the system turns raw price data into numbers: how strong each trend is, how cheap or expensive an asset is versus its own history, how much risk is in the market. It never forecasts — it scores what is already true.
02
Many signals, one verdict
No single measurement gets the final say — any one indicator is easy to fool. The system reads the market through several independent statistical lenses and blends them into one score, so the random noise in each cancels out and the real signal survives.
03
Score against a reference, then commit
A price means nothing on its own, so the system always compares. Kratos scores every crypto asset against the others and holds the strongest; Athena scores Bitcoin against its own long-run fair value and leans in when it is statistically cheap.
04
A gate that can step aside
Before risking capital, the system checks the overall market regime: is this an environment where the edge actually works? When conditions turn hostile it de-risks into defensive assets or cash. Knowing when not to play is half the edge.
05
Size the bet to the odds
Choosing what to hold is half the job; the rest is how much. Positions are sized to the edge and the risk on the table — larger when the signal is strong, smaller or nothing when the edge is thin or volatility spikes.
06
One calm decision a day
The whole pipeline runs once daily, on the close — measure, blend, score, gate, size, act. Feed it the same data and it returns the same answer every time: testable, repeatable, and free of the emotional mistakes that cost people money.
Test results
The test results — including the parts that hurt
Hiding drawdowns is what scammers do. Here is the whole curve: in-sample, out-of-sample, and every losing period, labeled.
Every figure in this section is a simulation. These are backtests run over historical
price data: no money was invested, no order was placed, and no customer has earned any of them. Each
chart says which kind it is — out-of-sample (settings chosen from past data only, then measured
on data the system had never seen), in-sample (settings chosen with hindsight), or a
full-history backtest. Trading costs are modelled asymmetrically and we would rather state it
than be asked: Kratos charges 0.30% on every unit of turnover, so its figures are already net of
trading cost — most published crypto backtests charge nothing. Athena models no fee and no slippage,
so its figures are gross. Neither models spread, taxes, execution failures or the cost of this subscription.
Kratos
Trend-following rotation · crypto majors + gold
out-of-sample labeled · losing periods shown
Kratos is a trend-following system: it holds the strongest asset available and steps aside when the trend breaks — every move made by rule, never emotion. In bull markets it rides the leading crypto major; when crypto weakens it rotates to gold or cash. It never predicts — it reacts to what the market is actually doing, once a day. Because it follows trends, it shines when markets move in clear directions and bleeds in choppy, sideways markets, by design. Aggressive: expect deep drawdowns as the price of the upside.
The honest test · walk-forward, out-of-sample · $10k, 2020–26Out-of-sample — settings chosen using only the data the system had at the time, then measured on data it had never seen. It is a simulation, not a record: positions are taken at the daily close and every unit of turnover is charged 0.30% in trading costs, which most published crypto backtests do not charge at all. Spread, slippage, taxes and the cost of this subscription are not modelled. No money was invested and nobody lived it.
Kratos (walk-forward)Bitcoin, bought & heldsimulated · out-of-sample $10k → $5.82MBTC held → $69.5kworst drop −49%
$9.5K$54.4K$311K$1.78M$10.14M
2020202220242026
$5.82M
Drawdown from peakworst −49.2%
0.0%−26.6%−53.1%
The in-sample ceiling · hindsight settings · $10k, 2020–26In-sample — the settings here were chosen with the whole period already visible, so this outcome was never achievable in advance. We publish it because the distance between this curve and the walk-forward above is the honest measure of how much of any backtest is hindsight. Same simulation rules, same 0.30% charged on every unit of turnover.
+58,077%Out-of-sample return — data the system never saw$10k → $5.82M vs $69.5k held in Bitcoin · simulated, after 0.30% turnover cost
+111,140%In-sample ceiling — settings picked with hindsight$10k → $11.1M · in-sample, never achievable in advance
−50%Drawdown to brace for−49% in the out-of-sample simulation
Bitcoin · bought & held
$69.5k$10k held, 2020–26+595%
Year by year, both simulated — the walk-forward (settings from past data only) and the in-sample ceiling (settings chosen with hindsight), each vs $10k held in Bitcoin.
Some years the honest test beats the ceiling — the optimized run picks one setting for all six years, not the best setting for each year. That is how you know neither is curve-fit to this table.
Athena
Bitcoin · mean-reversion accumulation & distribution
full-history backtest · no out-of-sample test · losing periods shown
Athena is a mean-reversion system: where Kratos buys strength, Athena buys value. It reads Bitcoin against a fair-value model and leans against the crowd. When the model flags BTC as statistically cheap, Athena accumulates harder; when it flags BTC as expensive, it trims and builds cash; the rest of the time it simply holds. That is the classic buy-low, trim-high discipline applied by rule instead of nerve, turning Bitcoin's violent swings into a lower average cost basis rather than a reason to panic. It does not try to call the exact top or bottom, use leverage, or promise you will beat the market. One honesty note: unlike Kratos's headline, which is out-of-sample, Athena's headline figure is a full-history backtest — the settings were chosen with the whole history visible, and it has no out-of-sample test behind it.
Full-history backtest · $10k, 2015–26 · Athena vs Bitcoin bought & heldA full-history backtest — not an out-of-sample test, and we label it that way rather than dress it up: the settings were chosen with all eleven years already visible. Unlike Kratos, this simulation charges no exchange fee and no slippage, so every Athena figure on this page is gross of trading costs. Its honest forward test is the public log, which records Athena's decision daily from the signal for 7 August 2026.
AthenaBitcoin, bought & heldsimulated · full-history backtest $10k → $16.36MBTC held → $2.0Mworst drop −59%
$7.1K$50.9K$365K$2.62M$18.82M
201520172019202120232025
$16.36M
Drawdown from peakworst −59.2%
0.0%−32.0%−63.9%
Athena
+163,521%Return — full-history backtest, no out-of-sample test$10k → $16.36M vs $2.0M held · simulated, gross of trading costs
−59%Worst drawdownholding BTC fell −83% over the same span
$33,103Average simulated buy pricethe model ends the simulation holding 224 BTC
Bitcoin · bought & held
$2.0M$10k held, 2015–26+19,977%
Over 4,231 backtest days — it buys cheap, trims expensive, and holds the middle.
Buy days997
Sell / trim days561
Hold (no trade)2,673
Peak simulated cash — model gains trimmed near highs, waiting to re-buy$2.2M
Past performance is not indicative of future results. Backtested and hypothetical results have inherent limitations and do not represent actual trading; hypothetical trading does not involve financial risk, and no hypothetical trading record can completely account for the impact of financial risk in actual trading. No money was invested, no order was placed, and Allocater has no live trading track record. All figures are Allocater's own, self-reported and not independently audited. Allocater is not an investment adviser, does not hold client funds, does not execute trades, and does not provide personalized investment advice. Crypto assets are volatile and you can lose money, including all of it.
In-sample = the data used to build a system; out-of-sample (walk-forward) = data it never saw — the honest measure of edge. Kratos's headline is its smaller out-of-sample number on purpose; the giant full-window figure is the in-sample ceiling, shown to disclose the gap between the best case and what a simulation actually produced. Athena's headline is a full-history backtest, labeled as such rather than dressed up as out-of-sample. What the simulations assume, stated plainly: positions are taken at the daily close, with no market impact. Kratos charges 0.30% (30 bps) of every unit of turnover, so its figures are already net of trading cost; its slippage setting is 0 bps and exists as a stress knob rather than an estimate. Athena models no fee and no slippage at all, so Athena's figures are gross — read the two accordingly. Neither models spread, funding, taxes, withdrawal delays or the cost of this subscription. Kratos's walk-forward window runs 6 July 2020 to 5 August 2026, chosen once and not re-picked; Athena's full-history backtest runs 1 January 2015 to 1 August 2026. Assets are excluded on any date before they were listed, so the systems never hold something that did not yet trade — but the roster itself (BTC, ETH, BNB, SOL, SUI, PAXG) was chosen by us in 2026, with the full price history of every one of them already visible. SUI did not exist before May 2023; SOL before 2020; BNB before 2017. A crypto rotation return of this size depends on the roster having contained the winners, and we picked the roster knowing who won. That is the single largest limitation of these figures, and we would rather state it than have you find it.
Read this before you buy
What quantitative investing is — and what it is not
These are not tips, calls, or day-trading. They are quantitative systems: every decision made by fixed rules and math, the same way every time. Our two systems are the two great families of systematic edge — Kratos follows trends (buying strength), Athena reverts to value (buying weakness) — designed to behave differently in the same weather. Before paying for either, it is worth understanding how they actually behave, including when and why they struggle.
01
This is quant investing, not day trading
A quant system makes every move by rule, not gut feeling. No reacting to news, no staring at charts, no "I feel like it'll go up." One calm update a day, executed the same way every time. Removing the emotional, impulsive human is the entire point — because that human is usually the problem.
02
Position size matters as much as position choice
Position sizing is half the game. Sized too large, a single rough stretch ends the exercise before the good years ever arrive. These systems size each position to survive the drawdowns on purpose — because surviving the bad patches is exactly what lets compounding do its work over time.
03
Timing matters — and the decision stays with the reader
The published results assume each day's position is taken near the daily close. In testing, a large share of the measured edge disappears when the same positions are taken "sometime tomorrow" instead — so a published result and a delayed one are not the same result. That is a property of the models, and it is why we publish the execution assumption rather than bury it. Whether to act on any published note, and when, is entirely the reader's own decision, and Allocater has no way to know whether anyone acted at all.
04
Trend-following — riding strength, not calling tops
Trend-following buys strength: it holds whatever is moving hardest, rides the trend while it runs, and steps aside when it breaks — no forecasts, no bottom-picking. It shines in clear, sustained moves — the big bull runs and the clean crashes it sidesteps — and struggles in choppy, sideways markets that whipsaw it in and out. That chop is the toll it pays to stay aboard for the next big move.
05
Mean reversion — buying what's cheap, not what's strong
Mean reversion buys weakness: it measures price against a fair-value model, accumulates when it is statistically cheap, trims when it is expensive, and holds the middle — leaning into fear, lightening into euphoria. It shines when price overshoots and snaps back toward fair value, and struggles when a runaway move stays cheap or expensive far longer than history says it should. Buying too early is the toll it pays.
06
No strategy works all the time — that is normal
Every edge has weather it loves and weather it hates — and the two families struggle at opposite times, which is exactly why we run both. A rough stretch is not the system breaking; it is the toll it pays to be there for the payoff it is built to catch. Expect deep drawdowns in both — a halving is within the tested range, not a malfunction, and nothing guarantees a recovery from one. Anyone who would abandon the rules at that point is not the right reader for this product.
07
Diversify the edge, not just the coins
Owning ten different coins is not diversification — in a crash they fall together. Real diversification means combining edges that profit in different weather. Trend-following and mean reversion are the classic pair: the stretch that punishes one often pays the other, so running both smooths the ride — shallower drawdowns, fewer white-knuckle months. It is the least-bad diversification available, and it is not a force field: in a true panic almost everything can fall at once. (The pricing card below explains how the two are published — separately, and never blended.)
08
The long game — a statistical edge, priced like an underwriter's
An insurer has no idea which house is going to burn down. It writes thousands of policies at prices set from measured frequencies, and it makes money — when it does — because the arithmetic holds across the whole book, not because any single policy was a smart bet. These systems work the same way: the edge is statistical, it shows up across full market cycles rather than over a week or a single trade, and no individual decision is ever meant to carry the argument. Short stretches are noise. And the honest half of the analogy is the half most people skip: underwriters lose money in catastrophe years, they know it going in, and they price for it. So do we. A market edge is estimated from history rather than fixed by a rulebook, which means it can decay, and it can be swamped by a bad stretch — which is precisely why every signal is forward-tracked in the open, where decay would show up in public.
Tested at scale — not one lucky backtest
A pretty backtest is easy to fake. This one was rebuilt and re-tested across twelve generations — over 2.3 million backtests, the largest single sweep covering 529,200 setting combinations across seven market eras, every sweep with its configuration count and its date recorded in our research log, and we will publish the underlying counts on request — and the honest walk-forward test re-chooses its settings from past data only, the way it would have had to be lived. Everything the data rejected is on record too, including three ideas we re-tested from scratch and rejected a second time.
Mid-development we found a bug in one of the system's indicators. We fixed it and re-ran every test from scratch — every number on this page is from the re-run. We also re-tested a setting that doubled the backtest return, then rejected it: on the honest test it performed worse. The backtest is not the product; the process is.
2.3 million+backtests across 12 generations
529,200settings · largest single sweep
7different market periods
past-data onlywalk-forward selection
The honest one-liner: these are disciplined processes for capturing big market trends and accumulating at good prices — not guarantees, not savings accounts, and not a way to get rich by next week.
Receipts, not reviews
Anyone can paste glowing quotes above a stock photo. We would rather show you things you can check yourself — proof that does not require taking anyone's word for it.
A public forward log, not screenshots
From the first recorded signal — 7 August 2026 for Athena, 9 August 2026 for Kratos — every daily decision is written to an append-only log and never edited afterwards, publishing once it ages past the two-week public delay. Rows dated earlier are labelled replays: today's engine re-run over past days, which proves nothing about what we said at the time. We would rather show a short honest record than a long flattering one. Nothing here asks you to take our word for it — it asks you to check. When accounts open, a free account will bring that to 3 days for Kratos and 7 for Athena, and subscribers will see each row live at the close.
Mid-development we found a bug in one of our indicators. We said so, fixed it, and re-ran every test from scratch. Ask a signal group when they last did that.
No fake reviews — ever
Manufactured social proof is where trust goes to die. Any testimonial you ever see on this page will carry a real name and a real, verifiable result — that is a standing promise, not a policy we will bend at launch.
The systems
Each is independent, with its own risk profile and time horizon. Both are published in full; which to read is the reader's choice.
Flagship · walk-forward tested
Kratos
A trend-following rotation system: it holds the single strongest major crypto asset, steps aside to gold or cash when the market turns, and never uses leverage. It thrives in trending markets and treads water in choppy ones. Aggressive and fully rules-based.
A mean-reversion system for Bitcoin — the mirror image of Kratos. A fair-value model flags when BTC is statistically cheap or expensive; Athena buys more near the lows, trims near the highs, and holds the middle. Across 11 years of Bitcoin history (2015–2026) that buy-low, trim-high discipline ended about 8× more than simply buying and holding Bitcoin, at a shallower drawdown (−59% vs −83%). Patient and fully rules-based.
risk · Patient · mean-reversionDaily signal · accum/dist
Two systems, built on opposite edges — because each one is weakest exactly where the other is strongest, and a single system, however good, leaves the whole book exposed to the one market it handles worst.
Opposite edges, on purpose
Kratos is a trend-following rotation: it wants persistence and does best when a move keeps going. Athena is mean reversion in a single asset: it wants stretch and does best when price has run far from its own centre of gravity. Trend-following and mean reversion are the classic pair because the conditions that punish one are usually the conditions that pay the other.
They cover different weather
A long grinding trend is where a reversion model gives ground — it keeps trimming into strength that does not stop. A violent snap back to the middle is where a trend model gives ground — it is still holding when the move ends. Neither problem is fixable inside the model that has it; the answer is a second model built on the other edge.
They are expected to disagree
On any given day the two can point in opposite directions, and that is the design working rather than a fault in it. Two systems that always agreed would be one system with extra steps. The disagreement is where the diversification actually lives.
Different assets, different clocks
Kratos rotates across a roster of major assets and rebalances daily. Athena reads one asset, Bitcoin, and most days does nothing at all. They are not two versions of the same bet at different speeds — they take different risks, on different things, on different schedules.
Published separately, never blended
Every figure published anywhere belongs to one system or the other. There is no blended backtest, no house split and no recommended allocation between them — Allocater knows nothing about any reader's circumstances and takes no view on them. Both are published in full; which to read, and what to do about either, is entirely the reader's own decision.
It is not a force field
Diversification across two edges is the least-bad protection available, not immunity. In a genuine panic almost everything can fall at once, and both of these can lose money at the same time. Each system's worst simulated drawdown is published beside its own curve, separately, for exactly this reason.
Two edges is a design decision about the models, not advice about anyone's portfolio. Nothing here is a recommendation to hold both, or either, or any particular amount of either.
Pricing
One subscription. Both systems.
There is one paid tier and it contains everything. The log stays free — a subscription removes the delay, nothing else.
Checkout is not open, and there is no price to show you — deliberately. Rather than announce a number and discount it later, Allocater is asking readers what the subscription should cost before fixing it. The poll below is anonymous: no email, no account, nothing stored in your browser, and no running tally shown — a visible tally would anchor every vote after it. When pricing is set it will be published here, with the reasoning, before checkout opens.
Help set the price
One paid tier with everything in it. No per-system pricing: splitting the systems invited buying whichever looked better lately, which is the performance-chasing both systems exist to prevent.
A monthly subscription that cancels anytime, and a one-payment lifetime option covering every system Allocater publishes later.
Planned risk reversal: a 30-day full refund on lifetime purchases, and monthly payments counting toward a later lifetime upgrade. Final policy publishes with the price.
The log stays free either way — a subscription removes the publication delay, nothing else.
Kratos and Athena were built as a pair, on opposite edges: the market that punishes one is usually the market that pays the other, which is why one subscription carries both rather than selling them as separate products. They are still published as two standalone models — every figure on this site belongs to one system or the other, and there is no blended backtest, no house split and no recommended allocation between them.
You keep custody · Monthly cancels anytime · Read the free log for as long as you like before you pay. View the live log →
The questions a skeptic asks
Is this a scam?
Fair question — you should ask it of everyone in this category. Our answer is structural: the logic is explained step by step, the backtests are labeled in-sample vs out-of-sample, drawdowns are shown beside every return, and from 9 August 2026 every signal is written to the public log and published there two weeks later — sooner with a free account, and live for subscribers, once accounts open. The log is append-only and dated, so count the rows yourself and decide what they are worth — we would rather you did that than take our word for anything. A business planning to deceive you does not publish its decisions before the outcome is known.
Do I hand you my funds?
No. You keep custody of your own assets on your own exchange or wallet — we send the daily signal, you place the trade. Allocater never holds funds and never trades for you.
What is the refund policy?
Lifetime purchases carry a 30-day full refund, no questions asked — enough time to watch the forward log and check the systems against it. Monthly subscriptions can be cancelled anytime and you keep access through the period you paid for, but monthly payments are not refunded. Neither option protects anyone from market losses — the refund covers the subscription price and nothing else. Checkout is not open yet, so neither policy is in force today; the forward log is free and public in the meantime.
What is the actual risk?
Real and material. Crypto and markets can lose value fast, and every system has losing periods — we show them on purpose. Position sizing is built in, but nothing here removes the risk of loss. Only allocate what you can afford to lose.
What if the system stops working?
Edges decay — we do not pretend otherwise. That is why we forward-track everything in the open, so degradation is visible to you in real time rather than hidden. If the public log shows the edge degrading, we will say so and pause new sales.
Do I need to be technical?
No. If you can place a market order on your exchange, you can run this. The dashboard tells you the current allocation; you place the trades on your own exchange.
Why should I trust the numbers?
You should not — you should verify them. The logic is explained step by step, out-of-sample results are labeled, and the forward log is public. Where a number is a backtest, we call it a backtest.
Risk disclosure
Investing in cryptocurrencies and other assets carries a high risk of loss, including the total loss of your capital. Prices are volatile and can move sharply against you.
Allocater is a daily research publication with a public, verifiable record — a bona fide publication of general and regular circulation, issued on a fixed daily schedule whether or not the models changed position. Every subscriber receives identical content. We collect no information about any reader's holdings, goals or circumstances, ask no suitability or risk questions, and are therefore structurally incapable of tailoring anything to anyone. Nothing published here is personalized financial, investment, legal or tax advice.
Allocater is not an investment adviser, a broker, an exchange, a custodian or an asset manager. It does not hold client funds, does not accept exchange API keys, does not execute trades, and does not provide personalized investment advice. Custody and control of any assets remain entirely with the reader, as does responsibility for every decision.
Everything Allocater publishes is designed, tested and published for spot, unleveraged positions only. It is not designed for and must not be used with futures, perpetual swaps, options, margin, or any leveraged product. No published figure is computed on a leveraged basis. Allocater publishes nothing about equities, ETFs, funds, or any other security.
Past performance is not indicative of future results. Backtested and hypothetical performance has inherent limitations: it is constructed with the benefit of hindsight, does not represent real trading, and does not account for all real-world costs and conditions. Out-of-sample and forward results are labeled as such.
Only risk capital you can afford to lose entirely. Consider seeking advice from a licensed professional before investing.
Do your due diligence. Then decide.
Read the methodology. Check the out-of-sample results. Watch the public forward log. If it holds up to your scrutiny — and we built it to — start when you are ready.
Lifetime: 30-day refund on the subscription price · Monthly: cancel anytime, no refunds · You keep custody
Watch it work before you decide
Join the list and we will send the weekly forward-log recap — what the systems signaled, and how it played out. When access opens you hear first, and the price will be published with its reasoning before checkout opens — it is being set with reader input now, in the anonymous poll above. No price has been set or charged yet, and nothing here is a contract.