Every closed level is measured against the capital at work when it closed, after fees, and compounded. Add them up and you get this number. That is the whole method, and it is public: Noema trades on our own capital on Binance spot and publishes the journal as it happens.
The index is the value of one unit of the strategy: 100 at launch, then priced every day with open positions at market. Below 100 means open levels are currently under their entry price; they are inventory waiting for their targets, not realized losses. Realized figures above come from closed trades only and are verifiable against exchange account history. How to read these numbers
Nothing is ever transferred to Oscillon. Your capital stays on your own Binance account, in your name, under your control. The algorithm connects through API keys that are allowed to trade and nothing else: they cannot withdraw, cannot transfer, cannot move funds anywhere. There is no Oscillon wallet, no deposit address and no step where you send money to us. You can revoke the keys on the exchange at any moment, and the algorithm simply stops.
One platform, one rule for all three: results are published here first, as they happen, before any of them is offered. Start with Noema, the one running on our capital today; the futures algorithms are further down, each with its own risk level stated plainly.
Spot grid on our own capital, published live.
What Noema is doing right now. The same number at four horizons: time-weighted profit from closed levels, after fees. Positions and order sizes are not published; the results of closing them are, every few minutes, straight from the journal.
All percentages: realized profit from closed levels, after fees, each level measured against the capital at work when it closed, compounded (time-weighted). Adding capital never dilutes them.
Each bar is the day's time-weighted profit from closed levels. Today is shown lighter while the day is still in progress.
| Closed | Market | Profit on the level's capital | Level was open |
|---|
Noema buys a market in levels and sells each level separately at its own target. Every row is one such level closed with profit: +0.50% means the capital committed to that one level (a small slice of the account) came back with half a percent on top after fees. This is the only place on the page where a percentage is not of the whole capital at work. Thousands of these per month add up to the figures above.
The risk profile of the Noema generation on a 0-100 scale: 0-25 low, 26-50 moderate, 51-75 elevated, 76-100 high. Computed only from measurable inputs: the execution journal, the state of the account and the nightly stress replays. Nothing here is an opinion; the full component table with its thresholds is in the monthly report.
| Month | Realized | Cumulative | Cycles | Days |
|---|
Per month, time-weighted. Months compound rather than add: the total is the product of the monthly factors, because each month's profit stayed at work in the next. The current month is marked as in progress.
Crash test. Every night the live configuration is re-run on the last four weeks of recorded market data with a historical BTC crash laid on top of it (May 2021, November 2022, August 2024). The figure is the marked-to-market dip of the budget in the worst of those scenarios: what this month would have looked like if that crash had happened inside it. It is a simulation, not a result.
Blocks. Market: drawdown, open share, worst days of the index. Stress: the crash test and how many markets the cash desk could carry to full depth. Structure: capital and markets at deep levels, concentration, deep cycles in 90 days. Liquidity: spread of held markets, number of venues. Operations: journal errors and unplanned restarts. A structural block above zero is by design for an accumulation strategy: deep levels are inventory waiting for price, and the index shows them rather than hiding them.
Profit that has already arrived on the account from closed levels, after fees. This is the headline figure. It is time-weighted: each closed level is measured against the capital at work at that moment and the results are compounded, so adding capital is never counted as profit and never dilutes past results. The share of all capital contributed to date is shown alongside for reference.
The value of one unit of the strategy, marked to market, starting at 100. It moves with the prices of open positions. Adding or withdrawing capital issues or redeems units at the current unit value, so the index does not jump when capital changes.
Noema buys in levels and sells each level only above its entry. Positions that have not yet reached their target are inventory: they show up as a dip in the index, not as a realized loss. Spot only, no leverage, no short exposure.
An illustration, not a forecast: your capital run through the realized record shown above, after exchange fees and after the platform fee of 20% on each month's net realized profit, with the plan minimum (99 USDT on Starter, 199 on Pro) applied in months where 20% comes to less. Part of the capital is set aside in BNB for fees, the way we do it on our own account. Pro runs Noema, the configuration behind the numbers on this page; Starter runs Eigen, the first generation with the base filter set.
| Basis | Pro · Noema | Starter · Eigen |
|---|---|---|
| Last 30 days, realized | - | - |
| Average month since launch | - | - |
| Last full month | - | - |
Noema is the configuration on our own account: the Eigen engine plus the selection rules fitted nightly in the lab. Eigen is the same engine with the base filter set; for this illustration it is assumed at 70% of Noema. An assumption, not a guarantee.
Rate-
| Horizon | Profit, USDT | Platform fee, USDT | Net to you, USDT | On capital |
|---|
Profit compounds because closed-level profit after the platform fee stays at work. The rate comes from this page: the realized result of the last 30 days or the average month since launch, time-weighted and net of exchange fees; for Starter it is scaled to Eigen. It changes every month, a month can be close to zero (then only the plan minimum applies), and a market-wide fall freezes capital in open levels for a time. Starter and Pro cover accounts up to 100,000 USDT. Past performance is not indicative of future results.
Computed by the same code that produces the monthly report, refreshed every few minutes. Left: the closed-trade record. Right: the marked-to-market unit index, the way a fund's NAV is measured. No amounts, only ratios and percentages.
| Closed levelsevery level is one buy and one sell with real fills | - |
| Profitable levelsa level closes only above its entry, so losses come from manual intervention or delisting only | - |
| Expected payoff per levelaverage net profit on the capital committed to a level | - |
| Profit factor (net)gross profit divided by gross loss; shown once there are at least 5 losing levels, otherwise not meaningful | - |
| Levels per dayaverage over the live period | - |
| Median holding timefrom a level's buy to its sell | - |
| Longest winning streakconsecutive profitable levels | - |
| Max open orderslevels held simultaneously at the peak | - |
| Time-weighted return (TWR)change of one unit since launch, deposits and withdrawals excluded | - |
| Maximum drawdownlargest fall of the unit from its peak; temporary, since levels below entry are inventory | - |
| Sharpe ratioannualized, risk-free rate 0, on daily unit returns; needs 30 days of data | - |
| Sortino ratioannualized, penalizes only downside days | - |
| Calmar ratioannualized return divided by maximum drawdown | - |
| Days in sampledaily observations behind the ratios above | - |
These ratios describe the market value of one unit, which moves with the prices of open positions even on days when nothing is sold. A negative day here means open levels were priced lower that day, not that anything was sold at a loss; realized results are in the left column and never include a losing day by construction.
One long-only accumulation engine, two generations. It buys structured levels into weakness, sells each level at its own target, and only trades markets where that behaviour has a measured edge. Eigen trades with the base market filters; Noema adds the selection rules fitted nightly in the lab. The grid is the simple part. The edge is in market selection, capital protection, and the discipline of measuring every change before it goes live.
Each generation is a distinct algorithm with its own name and its own track record, on one research and execution platform. Eigen, the first, is the base. Noema, the second, runs on our own capital and is the record on this page. Quanta is being tested in the lab now, will go live next and will be available on Pro. The tabs in Live analytics show all three side by side.
Long-only structured accumulation on spot markets: the level ladder with its rescue tail, the shared capital desk with crash coverage, and the base market filters. The engine every later generation runs on.
The Eigen engine with learned market selection: veto families, hourly rules and a market memory fitted nightly in the selection lab on every USDT market. Running on our own capital; the record on this page is Noema's.
Quanta reads the state of the market before the engine acts on it and splits capital into discrete allocations across generations, markets and venues: when to accumulate, when to slow down, when to hold cash and when to hedge the book, rebalanced by measured expected value.
Futures grid with a stop on the exchange for every position.
Axiom is our futures algorithm, now in development. Before it touches real capital it trades a Binance demo futures account with real exchange orders, right here: leverage capped at 5x, a hard stop resting on the exchange for every position, so it holds even if our software stalls, and a monthly loss limit after which it stops until the next month. Losses are shown as they happen. It is separate from Noema, and no client capital is involved.
How risky the Axiom demo account is right now, on the same 0-100 scale as Noema: 0-25 low, 26-50 moderate, 51-75 elevated, 76-100 high. Computed from the live account state, the exchange journal and the replay on history.
Market: share of the monthly loss limit already used, largest drop of the account, Bitcoin below its 14-day average and its fall over 24 hours. Exposure: worst case of all open cycles closing at their stops against what is left of the monthly limit, positions without a stop, the gap between each stop and its liquidation price, how deep the grids are, markets of one sector. Stress: an instant 10% gap on every market that jumps over the stops, and the worst month of the replay on history. Operations: exchange errors in the journal and stops the exchange did not accept, held by our software instead.
Axiom trades only coins from a fixed list of large, liquid markets. Each day it keeps the ones whose price has recently kept returning to its average and drops those in a strong trend.
| Market | Entry | Price | Stop on exchange | Open P&L |
|---|
Event-driven replay on 5-minute futures candles with fees, funding, stop slippage and simulated two-hour outages of our own software during sharp falls. "As originally designed" is the same grid at 20x leverage with no stop, the way such strategies are commonly run.
Demo-account trading is research, not a forecast or a promise of returns. Demo order books are thinner than the live market, so real execution will differ. Axiom reaches real capital only after several months of results that hold up.
High risk, high return, for those who like it rough.
Flux is the opposite of Axiom by design: a deep averaging grid with no stop-loss, 10x leverage, grid exposure of three times the capital, and a counter-trend scalper that feeds margin to a position stuck underwater. A single market can be liquidated - that is accepted on purpose and shown here as it happens. The only brakes: a monthly loss limit of 25% after which it stops until the 1st, and a 30-day ban on any market that got liquidated. It trades a separate Binance demo account; no client capital is involved.
How risky the Flux demo account is right now, on the same 0-100 scale as Noema and Axiom. Expect it to sit higher than Axiom's: no stops, 10x and a grid three times the capital are built into the design.
| Leverage | Grid exposure | Liquidation from average | Result | Liquidations |
|---|
You choose the speed. Leverage does not change what a trade earns at the same size; it changes how large the grid is for the same margin and how close liquidation sits. The table replays Flux's own closed cycles as if the leverage had been different: a cycle whose worst excursion from the average price crossed that leverage's liquidation line counts as liquidated. Higher rows earn more in calm markets and lose the whole position sooner in a crash.
| Market | Entry | Price | Liquidation | Open P&L |
|---|
Flux is in demo testing. It reaches real capital only on a separate account, with the client's explicit acknowledgement of the risk, and only after its demo record - including its worst month - has been published here for long enough to judge.
Eigen and Noema run on a research and execution platform built to carry several algorithms. The platform records the market, replays it, fits selection rules on it, runs the live engine and reports the result, every night, on its own.
Trades, best bid and ask and candidate-market ticks recorded continuously.
The live engine re-run on recorded ticks with a shared desk, real fees and crash overlays.
Features of every USDT market at each decision day against its 30-day outcome.
Execution, selection and capital desk in production, with shadow variants running alongside.
Investor reports and this public feed from one journal, one set of formulas.
Subscriptions open soon, by referral code. The code opens your account; the plan is chosen and paid inside it, and the instance is provisioned by us right after the payment. Your capital never leaves your exchange account: the algorithm trades it through API keys that can trade but cannot withdraw, from a dedicated instance provisioned for you.
Email, password and a code from a partner who already runs Eigen or Noema. Confirm your email from the letter and your account is open. No code yet? Register your interest and we invite you when a place opens.
Starter or Pro: the plan minimum in USDT, in advance; from then on each month is invoiced as 20% of that month's net realized profit, or the minimum, whichever is larger. Your instance is prepared automatically as soon as the payment is confirmed, usually within the hour. Institutional accounts open with a second, institutional code issued once the terms are agreed: no invoice, the Noema instance is prepared right away.
Create API keys on your Binance account with trading permission only, no withdrawals, restricted to the IP address shown in your account, and enter them in your instance. Then switch trading on.
Realized result, open positions, assets and open orders, the profit calendar and the index for your account, in the same accounting used on this page. Pro switches between Eigen and Noema in one click; Institutional sees the fee accrual live.
Oscillon never holds client funds. Keys are stored encrypted and can be revoked on the exchange at any time.
Client instances run the production release of the engine, Eigen or Noema by plan, and report with the same unit accounting shown on this page.
Ladder depth and reserve are derived from the account balance, the same way they are for our own capital.
Launching soon, by referral code only. Your capital stays on your exchange account. Oscillon takes 20% of the net realized profit each month, with a plan minimum that covers access, the instance and the dashboard in months where 20% comes to less. Results are published on this page, so you judge before you pay, and every month after. Two plans for accounts up to 100,000 USDT; above that, an individual agreement.
Run a site, a channel or an agency with an audience that trades? Register as a partner and get a personal link. Every client who comes through it is attributed to you for as long as they stay, and you receive 25% of the platform fee we collect from them: 5 of our 20 points, paid monthly in USDT. The link does not expire; you see your clients, the fee accruing and your share in your own account.
An illustration, not a forecast. Each client is assumed on Pro (Noema) at the realized rate shown on this page; the platform fee is 20% of each month's net realized profit, never less than the plan minimum of 199 USDT, and your share is a quarter of it. A month can be close to zero, then your share is a quarter of the minimum. Past results are not a forecast.
An independent algorithmic trading research company founded by Sergei Ferov. We build the platform, run our own capital on the algorithms first, and publish the results as they happen.
For the full monthly report, a referral, or a live walkthrough of the research platform:
hello@oscillon.io