Confluence Toolkit View reports
VIEN

System running · data updates automatically

A quantitative trading system for perpetual futures markets

A quantitative research and trading system built by Uniconn, running on perpetual futures markets. This page publishes the raw per-instrument backtest results — every row links to the full report, nothing withheld.

View 120 backtest reports Roadmap →
Principles

What we are building

A trading system that can be verified with numbers, rather than one that rests on intuition or a single indicator.

Evidence before deployment

Every trading parameter must survive walk-forward out-of-sample validation before it is used. No figure on this page comes from in-sample optimisation.

The right yardstick

Win rate is not the target. The system is judged on expectancy (expected return per trade, measured in R) and profit factor.

Real costs, charged properly

Trading fees and funding are charged to each backtested trade the way the exchange actually bills them — not deducted as an estimate at the end.

Automated end to end

From data collection, backtesting and per-instrument parameter fitting through to 24/7 market monitoring and alerting — running without manual intervention.

Validation scale

Scale of testing

Read directly from the system's calibration database at the time this page was built.

Instruments fitted182
Long/short configs367
Trades backtested10,040
Reports published120
Candles analysed3,653,598
Average win rate52%
Coins fitted & reports22/0802/09
Average win rate29/0802/09
Candles analysed29/0802/09
Trades backtested22/0802/09

Trend per deployment (deploy log below). Instruments fitted Reports published

Raw data

Backtest results per instrument

121 results across 120 instruments, timeframes: 15m, 4H. Each timeframe is its own row — results are not directly comparable across timeframes, because a different stop distance means trading costs consume a different share of the edge. Click "Open report" for the source report: price chart, every entry and exit, equity curve, and a before/after comparison of the fitting step.

Read this before interpreting any number below.
  • These are backtest results, not live trading performance. The system currently runs in simulation; no real-money results have been published.
  • A “gross” label means fees and funding have not been deducted. Cost accounting was added recently and 122 of 367 configs are still awaiting re-measurement. On the first config measured with costs charged, profit factor fell markedly against its gross figure — so treat every “gross” number as an upper bound, not an expected result.
  • n < 30 (amber) is too small a sample to conclude anything. The system does not trade real money on configs below that threshold.
Instrument TFCandles LONGSHORT CostsReport
nWinPFExp (R) nWinPFExp (R)
AAVEAAVE-USDT-SWAP15m34.560944%0.80-0.0742241%1.06+0.037netOpen report ↗
ADAADA-USDT-SWAP15m34.560956%1.43+0.20921349%1.03+0.017netOpen report ↗
ALGOALGO-USDT-SWAP15m34.5604151%1.24+0.1213850%1.04+0.020netOpen report ↗
ALLOALLO-USDT-SWAP15m17.2801258%1.61+0.1854351%0.80-0.105netOpen report ↗
APRAPR-USDT-SWAP15m29.1631457%0.92-0.0254455%1.32+0.149netOpen report ↗
APTAPT-USDT-SWAP15m34.5601155%1.24+0.0771953%1.05+0.024netOpen report ↗
ARBARB-USDT-SWAP15m34.5601958%1.23+0.0931050%1.02+0.009netOpen report ↗
ARKMARKM-USDT-SWAP15m34.5602352%1.34+0.1701464%1.94+0.217netOpen report ↗
ASTERASTER-USDT-SWAP15m31.4921242%1.91+0.3392748%0.80-0.095netOpen report ↗
ATHATH-USDT-SWAP15m34.5601947%0.69-0.1644856%1.28+0.127netOpen report ↗
ATOMATOM-USDT-SWAP15m34.5601833%0.43-0.3501856%1.76+0.338netOpen report ↗
AVAXAVAX-USDT-SWAP15m34.5602264%1.58+0.2449559%1.13+0.055netOpen report ↗
BABYBABY-USDT-SWAP15m34.5602650%0.92-0.04318040%0.64-0.239netOpen report ↗
BCHBCH-USDT-SWAP15m34.5602433%0.53-0.2752454%1.51+0.222netOpen report ↗
BEATBEAT-USDT-SWAP15m26.6694168%1.74+0.2264159%1.02+0.008netOpen report ↗
BERABERA-USDT-SWAP15m34.5602050%1.28+0.12616345%0.80-0.116netOpen report ↗
BICOBICO-USDT-SWAP15m34.5601753%1.96+0.3831346%1.74+0.308netOpen report ↗
BILLBILL-USDT-SWAP15m10.6132642%1.00+0.0002544%0.83-0.100netOpen report ↗
BIOBIO-USDT-SWAP15m34.5601050%1.22+0.0999851%1.15+0.074netOpen report ↗
BNBBNB-USDT-SWAP15m34.5604243%0.67-0.1761527%0.38-0.368netOpen report ↗
BOMEBOME-USDT-SWAP15m34.5603348%0.96-0.02326144%0.93-0.043netOpen report ↗
BRETTBRETT-USDT-SWAP15m34.5601567%1.75+0.2768952%1.16+0.077netOpen report ↗
BTCBTC-USDT-SWAP15m34.5606134%0.61-0.2061242%0.62-0.196netOpen report ↗
CCCC-USDT-SWAP15m27.6847449%1.01+0.00312243%0.75-0.136netOpen report ↗
CHIPCHIP-USDT-SWAP15m11.795944%0.73-0.1681250%0.88-0.064netOpen report ↗
COAICOAI-USDT-SWAP15m17.2803949%1.11+0.064944%0.51-0.295netOpen report ↗
CROCRO-USDT-SWAP15m34.5601547%0.76-0.1371070%1.41+0.100netOpen report ↗
DATADATA-USDT-SWAP15m4.8815941%0.55-0.298757%1.87+0.400netOpen report ↗
DOGEDOGE-USDT-SWAP15m34.5602343%0.85-0.0882854%1.33+0.141netOpen report ↗
DOSDOS-USDT-SWAP15m1.381540%0.55-0.286933%0.12-0.619netOpen report ↗
DOTDOT-USDT-SWAP15m34.5606545%0.70-0.16315546%0.80-0.112netOpen report ↗
DYDXDYDX-USDT-SWAP15m34.560944%1.38+0.1981547%1.18+0.092netOpen report ↗
EGLDEGLD-USDT-SWAP15m34.5604842%0.49-0.339967%1.63+0.231netOpen report ↗
ENAENA-USDT-SWAP15m31.9891338%0.90-0.071944%2.78+0.438netOpen report ↗
ENJENJ-USDT-SWAP15m34.5601346%0.91-0.0482259%1.60+0.248netOpen report ↗
ESPESP-USDT-SWAP15m18.099967%3.38+0.624956%0.78-0.102netOpen report ↗
ETCETC-USDT-SWAP15m34.5601464%5.76+0.8152756%1.38+0.145netOpen report ↗
ETHETH-USDT-SWAP15m34.5601547%0.63-0.1991155%2.30+0.325netOpen report ↗
FILFIL-USDT-SWAP15m34.5604048%1.00+0.0011369%3.61+0.674netOpen report ↗
FLOWFLOW-USDT-SWAP15m34.5603345%0.89-0.0602148%1.04+0.022netOpen report ↗
GALAGALA-USDT-SWAP15m34.5601070%1.26+0.0901867%1.04+0.017netOpen report ↗
GASGAS-USDT-SWAP15m34.560757%0.95-0.0162650%1.54+0.266netOpen report ↗
GMXGMX-USDT-SWAP15m34.5602348%1.36+0.2011839%1.15+0.076netOpen report ↗
GRAMGRAM-USDT-SWAP15m5.7162259%1.18+0.0672552%1.73+0.390netOpen report ↗
HBARHBAR-USDT-SWAP15m34.5601323%0.34-0.464650%2.64+0.614netOpen report ↗
HYPEHYPE-USDT-SWAP15m34.5603352%1.16+0.0721765%1.98+0.395netOpen report ↗
ICPICP-USDT-SWAP15m34.5601164%1.39+0.1211338%1.40+0.179netOpen report ↗
IOTAIOTA-USDT-SWAP15m34.5603043%0.62-0.1971275%5.16+0.535netOpen report ↗
JELLYJELLYJELLYJELLY-USDT-SWAP15m34.5602839%0.58-0.24025050%0.96-0.021netOpen report ↗
JUPJUP-USDT-SWAP15m34.5603951%1.14+0.0652442%0.87-0.064netOpen report ↗
KAITOKAITO-USDT-SWAP15m34.5607100%—+0.9711250%1.47+0.260netOpen report ↗
KSMKSM-USDT-SWAP15m34.5601457%0.83-0.0821070%2.67+0.571netOpen report ↗
LABLAB-USDT-SWAP15m28.2161953%0.77-0.1181362%1.20+0.080netOpen report ↗
LINEALINEA-USDT-SWAP15m34.2141471%3.11+0.3453234%0.30-0.448netOpen report ↗
LINKLINK-USDT-SWAP15m34.5602343%0.91-0.051580%1.86+0.187netOpen report ↗
LITLIT-USDT-SWAP15m22.953967%4.10+0.4892669%1.26+0.086netOpen report ↗
LPTLPT-USDT-SWAP15m34.5601527%0.76-0.1712650%0.67-0.146netOpen report ↗
LTCLTC-USDT-SWAP15m34.5602756%1.88+0.2392642%0.70-0.194netOpen report ↗
MEGAMEGA-USDT-SWAP15m11.193771%1.81+0.259757%2.00+0.479netOpen report ↗
MEMEMEME-USDT-SWAP15m34.5602060%1.54+0.2111553%1.15+0.058netOpen report ↗
MINAMINA-USDT-SWAP15m34.5601958%2.02+0.3721457%1.83+0.334netOpen report ↗
MINIMAXMINIMAX-USDT-SWAP15m5.462540%1.63+0.1351155%1.04+0.015netOpen report ↗
MOODENGMOODENG-USDT-SWAP15m34.5601127%0.39-0.4331443%0.67-0.182netOpen report ↗
MORPHOMORPHO-USDT-SWAP15m34.5604365%2.23+0.4441030%0.64-0.269netOpen report ↗
NEARNEAR-USDT-SWAP15m34.5602433%0.38-0.4051631%0.66-0.242netOpen report ↗
NEIRONEIRO-USDT-SWAP15m34.5605050%0.92-0.0431275%2.93+0.576netOpen report ↗
NMRNMR-USDT-SWAP15m34.5602560%1.71+0.2301833%0.47-0.304netOpen report ↗
OO-USDT-SWAP15m6.1181242%0.56-0.2876050%0.93-0.034netOpen report ↗
OKBOKB-USDT-SWAP15m33.2221242%0.78-0.0891457%2.13+0.302netOpen report ↗
ONDOONDO-USDT-SWAP15m34.5603551%0.87-0.0601136%1.45+0.138netOpen report ↗
ONTONT-USDT-SWAP15m34.5601650%0.82-0.08014555%1.20+0.083netOpen report ↗
ORDIORDI-USDT-SWAP15m34.5603256%1.16+0.068757%1.44+0.132netOpen report ↗
PEOPLEPEOPLE-USDT-SWAP15m34.5601963%1.17+0.0721275%2.72+0.454netOpen report ↗
PEPEPEPE-USDT-SWAP15m34.5608046%0.72-0.1561338%0.30-0.439netOpen report ↗
PIPI-USDT-SWAP15m34.5601553%0.79-0.11621446%0.94-0.032netOpen report ↗
PLUMEPLUME-USDT-SWAP15m34.5606052%1.19+0.0922446%1.51+0.289netOpen report ↗
PNUTPNUT-USDT-SWAP15m34.5603244%1.05+0.0281457%2.41+0.454netOpen report ↗
POLPOL-USDT-SWAP15m34.5605143%0.71-0.1671040%0.37-0.443netOpen report ↗
POPCATPOPCAT-USDT-SWAP15m34.560967%2.48+0.3772060%2.01+0.296netOpen report ↗
PROSPROS-USDT-SWAP15m11.3063569%2.76+0.4961354%2.09+0.281netOpen report ↗
PUMPPUMP-USDT-SWAP15m34.5601844%0.89-0.0632357%1.49+0.206netOpen report ↗
QNTQNT-USDT-SWAP15m6.9214955%1.17+0.0702854%1.16+0.078netOpen report ↗
QTUMQTUM-USDT-SWAP15m34.5603855%1.55+0.2061182%3.97+0.341netOpen report ↗
RAVERAVE-USDT-SWAP15m23.9636048%0.69-0.1698748%1.00-0.000netOpen report ↗
RERE-USDT-SWAP15m6.2921233%0.72-0.1545237%0.60-0.267netOpen report ↗
RENDERRENDER-USDT-SWAP15m34.5601567%1.33+0.1223345%1.01+0.003netOpen report ↗
RIVERRIVER-USDT-SWAP15m21.7852955%1.04+0.019838%0.99-0.007netOpen report ↗
RSRRSR-USDT-SWAP15m34.5602050%1.01+0.0062836%0.92-0.053netOpen report ↗
RVNRVN-USDT-SWAP15m34.5601741%0.58-0.2961354%1.37+0.130netOpen report ↗
SHIBSHIB-USDT-SWAP15m34.5602357%1.19+0.0861040%0.55-0.250netOpen report ↗
SKYSKY-USDT-SWAP15m32.5353441%0.80-0.1301856%1.01+0.003netOpen report ↗
SOLSOL-USDT-SWAP15m34.56010845%0.95-0.0311050%0.91-0.023netOpen report ↗
SOONSOON-USDT-SWAP15m34.560786%3.81+0.4751958%0.81-0.079netOpen report ↗
SPKSPK-USDT-SWAP15m34.5601354%1.15+0.0651560%2.02+0.316netOpen report ↗
STABLESTABLE-USDT-SWAP15m24.0981164%2.20+0.502944%1.03+0.013netOpen report ↗
STRKSTRK-USDT-SWAP15m34.5606057%1.47+0.1882050%0.93-0.038netOpen report ↗
SUISUI-USDT-SWAP15m34.5601267%2.71+0.3392273%2.71+0.455netOpen report ↗
SUSHISUSHI-USDT-SWAP15m34.5602138%0.91-0.0372548%1.25+0.113netOpen report ↗
TAOTAO-USDT-SWAP15m34.56011147%0.74-0.1291547%1.04+0.022netOpen report ↗
TRBTRB-USDT-SWAP15m34.5602152%1.27+0.1171540%0.42-0.395netOpen report ↗
TRXTRX-USDT-SWAP15m34.5602934%0.50-0.3532040%0.54-0.239netOpen report ↗
TURBOTURBO-USDT-SWAP15m34.5601155%1.27+0.1141527%0.54-0.319netOpen report ↗
UNIUNI-USDT-SWAP15m34.5604556%0.98-0.0092741%0.61-0.250netOpen report ↗
UPUP-USDT-SWAP15m13.0111457%0.93-0.033944%0.66-0.200netOpen report ↗
USELESSUSELESS-USDT-SWAP15m34.5604862%1.84+0.3361145%1.26+0.124netOpen report ↗
VVVVVV-USDT-SWAP15m4.2674461%2.08+0.406560%1.02+0.010netOpen report ↗
WW-USDT-SWAP15m34.560743%1.82+0.2779551%1.01+0.006netOpen report ↗
WALWAL-USDT-SWAP15m34.5601136%0.51-0.3552148%1.58+0.222netOpen report ↗
WLDWLD-USDT-SWAP15m34.5604648%0.86-0.0651436%0.43-0.318netOpen report ↗
WLFIWLFI-USDT-SWAP15m34.3387647%0.88-0.0592138%0.48-0.280netOpen report ↗
XLMXLM-USDT-SWAP15m34.560786%2.05+0.174862%1.05+0.022netOpen report ↗
XPLXPL-USDT-SWAP15m34.5601369%2.38+0.3542050%1.27+0.135netOpen report ↗
XRPXRP-USDT-SWAP15m34.5601145%0.32-0.3711547%1.19+0.087netOpen report ↗
YFIYFI-USDT-SWAP15m34.5601060%5.24+0.4861638%0.60-0.203netOpen report ↗
ZECZEC-USDT-SWAP15m28.2331656%1.07+0.0351464%1.61+0.143netOpen report ↗
ZENZEN-USDT-SWAP15m27.9891464%1.87+0.3491155%1.37+0.152netOpen report ↗
ZETAZETA-USDT-SWAP15m34.5605730%0.57-0.3321839%0.60-0.242netOpen report ↗
ZILZIL-USDT-SWAP15m34.5601233%0.61-0.189825%0.27-0.334netOpen report ↗
ZKZK-USDT-SWAP15m34.5601267%2.94+0.5631060%1.60+0.272netOpen report ↗
ZRXZRX-USDT-SWAP15m34.5601155%1.47+0.1982050%1.15+0.045netOpen report ↗
ETHETH-USDT-SWAP4H2.160not fitted667%3.22+0.762netnone

n — trades in the sample. Win — share of winning trades. PF — profit factor (gross profit ÷ gross loss; >1 means profitable within the sample). Exp — expectancy, expected return per trade in R (1R = the loss taken if the stop is hit). Expectancy is the measure that compares against trading costs; profit factor does not.

Note: the linked reports are currently published in Vietnamese only. The charts, trade tables and metrics are readable regardless of language.

Shipped controls

Security & operational discipline

What follows lists only mechanisms already implemented in the running code — not intended policy. Each item maps to a specific gate in the source; what is still missing is stated plainly at the end of the section.

Control access 2

The bot only listens to authorised operators

Commands from accounts outside the authorised list are dropped with NO reply — a prober gets no signal that they reached the right bot at all.

Simulation keys and real-money keys are physically separate

The exchange issues two different API key pairs for the two environments. The key running the simulation technically CANNOT place a single real order — not because a config flag says so, but because the exchange rejects it.

Pre-trade gates 4

Trades nothing by default

The master switch defaults OFF and the allow-list starts EMPTY. No coin/side pair trades until a human explicitly opts that exact pair in.

Configurations without full cost accounting are refused

Any configuration measured before fee accounting existed is blocked outright. The measured reason: of 16 pairs that had passed every other gate, 38% fell to PF<1 the moment round-trip fees were charged.

Real-money threshold: n≥30 and PF>1, net of costs

Too small a sample, or no edge after fees, and the bot refuses to add the pair — even when the operator explicitly commands it. This is a gate in code, not a habit of self-discipline.

The trading signal grade is hard-pinned

The signal-quality bar for entering a trade is pinned in code. It never inherits the alerting threshold, which an operator may have loosened for an unrelated purpose.

Loss limits 3

Circuit breaker on daily realised loss

Hitting the daily realised-loss limit stops NEW entries until the next day. Crucially it does NOT stop management of already-open positions — abandoning defence midway is more dangerous than having no breaker at all.

The stop-loss lives on the exchange, not in the bot

The stop is placed on the exchange as a server-side conditional order at entry time. The position stays protected even if the machine running the bot loses power or the process dies between scan cycles.

Undersized trades are skipped, never inflated

When the risk budget yields a size below the exchange minimum, the system SKIPS the signal. It never rounds up to become placeable — rounding up is quietly trading above the risk limit you set.

One control is still MISSING.

The safety model currently rests on position size and per-trade risk budget, NOT on a leverage limit — there is no hard leverage ceiling in the code. At high leverage the margin buffer can be thin enough that a position is liquidated before the stop-loss triggers, which defeats the strongest control listed above. This sits in the roadmap below and is not done.

Progress transparency

Roadmap & technical debt

The project publishes what it has fixed alongside what it still owes. Every deployment updates this list, and the log below is written by the system itself — not self-reported.

In progress3
Planned7
Done7
In progress 3
In progressResearch

Mapping the full space of not-yet-fitted parameters

Overall results (win rate, PF, expectancy) are not yet favourable, prompting a full review of every system parameter — from signal scoring down to entry/exit — to separate what has actually been fitted from what is still hardcoded. Finding: the per-condition scoring weights, the stop-loss buffer, and the tiered take-profit levels have never been fitted — the stop-loss buffer in particular feeds directly into fees eating the edge on short timeframes (a cause already measured elsewhere). Proposed trial order: (1) stop-loss buffer + take-profit levels — medium effort, addresses the known fee problem head-on; (2) switch the fitting objective to one with a statistical-significance check instead of raw risk-adjusted return; (3) real 1H/4H fitting once enough data exists; (4) signal condition weights — saved for last, the highest overfitting risk on the whole list.

Why it mattersMapping the parameter space before touching it, to know what's actually being optimized and where the overfitting risk sits — nothing on this list is implemented yet.

In progressTrading costs

Re-fitting every config with costs charged

Every config produced before cost accounting existed is a gross figure. They are being re-run batch by batch, in parallel across multiple machines — currently at 216/400 configs (54%), up from the initial 12% (41/346). What's left: keep running the remaining instruments, merge results from parallel machines without losing data, then rebuild the corresponding reports.

Why it mattersThis is a precondition for any discussion of real-money trading.

In progressResearch

Learning stop-loss/take-profit — extending to every alert, not just real-money trades

The mechanism to learn from the maximum adverse/favourable excursion of each closed trade is built, validates out-of-sample, and never applies anything without human confirmation — but it sits idle because too few REAL-MONEY trades exist to learn from (only a small set of coin/side pairs is cleared for real trading, so most alerts never produce a real trade). Expanding the data source: simulating the outcome of every alert fired using the exact same stop-loss/take-profit rules the live system uses, reusing the existing mechanism on a much larger dataset instead of building a parallel one. Simulated results are clearly labelled apart from real trades (more optimistic since slippage is ignored), and the project's n≥30 bar still applies before any analysis — below that, results are only logged, no conclusions drawn.

Why it mattersIf it works, this is the most realistic path to answering with data whether the system is currently cutting winners short — there currently isn't enough data to say.

Planned 7
PlannedTrading costs

Switching to resting orders to cut fees

Resting orders carry a materially lower fee than orders that cross the spread, at the cost of not always being filled. The real fill rate has to be measured before this can be adopted.

Why it mattersDirectly reduces the cost charged to every trade.

PlannedTrading costs

Testing higher timeframes so costs stop consuming the edge

Already measured: the short side is net profitable on both the hourly (+0.098R, PF 1.70) and 4-hour (+0.218R, PF 2.20) timeframes, while the long side still loses on every timeframe. But fitting each instrument separately at these timeframes is blocked by sample size — 12 months of data gives a median of only ~12 trades/instrument at 1H, ~3 at 4H, far short of the n≥30 bar. Two directions under consideration: pooling signals across coins on the same timeframe to reach a usable sample for one shared config (trading away per-coin specificity), or extending history beyond the current 12 months. This is logged technical debt with no set deadline — not work currently in progress.

Why it mattersIf the sample-size problem is solved, this addresses the cost problem at its root — but it depends on data, not just code.

PlannedResearch

Machine-learning model for entry detection

Replacing the hand-tuned scoring system with a model learned from labelled data. Currently at the feature-engineering and trade-labelling stage on a pilot instrument. It must beat the current system out-of-sample before it replaces anything.

Why it mattersA long-term direction; no result yet strong enough to replace the current system.

PlannedRisk control

Finding: most watchlist alerts rest on an unreliable sample size

Traced back to a real alert: a coin lost three consecutive trades in the direction the bot flagged. Investigation showed the bot behaved exactly as designed — the problem was the quality of that coin's fitted configuration (n=13, far below the project's own n≥30 bar). Scanning the full watchlist (182 coins) found 161 coin/side pairs currently capable of triggering an alert despite n<30, 58 of which also hit a walk-forward scoring error from having too few trades to split into stable validation folds — the same failure pattern, not an isolated case. This is a real gap between a stated principle (don't trust n<30 results) and enforcing that principle at the actual point where an alert reaches a person — today the gate only exists before real-money trading, not before a watchlist/demo alert.

Why it mattersHigh priority — directly affects the experience and trust of anyone following alerts, even before any real money is involved.

PlannedResearch

Separating market regimes during fitting

Fitting across mixed market regimes dilutes the result. Each config should represent a single regime.

Why it mattersRaises confidence in fitted parameters and reduces overfitting risk.

PlannedInfrastructure

Automating calibration sync between server and repository

Currently synced by hand after each fitting run on the server. This is technical debt with a risk of the two copies drifting apart.

Why it mattersReduces operational risk; no effect on trading results yet.

PlannedValidation discipline

Validating a proxy indicator against source data

One internal indicator is currently inferred indirectly from candle data. An initial check suggests it agrees in direction with the source trade data only about a third of the time, on a small sample. The sample needs widening, and the proxy replacing with source data if that holds.

Why it mattersA weak indicator trusted too far drags down every condition that depends on it.

Done 7
DoneValidation discipline

Blocked pre-cost configs from real-money trading

The system now refuses any config that has not been re-measured with costs charged, however good its old numbers look.

Why it mattersAn optimistic figure can no longer reach real money by accident.

DoneRisk control

Sizing capped by margin, demo and real money fully separated

The order-size cap used to bound position notional — at high leverage the actual margin committed could be far smaller than it looked. The cap now bounds margin directly, and since the resulting notional can then exceed what a single market order accepts, the system automatically splits into smaller orders when the exchange rejects one. Demo and real-money caps are now also fully separate — turning on real trading no longer silently inherits a limit meant for aggressive demo testing.

Why it mattersNarrows the gap between intended risk and actual margin risk — especially relevant before any consideration of enabling real-money trading.

DoneRisk control

Capping leverage so the stop-loss can always fire

At maximum leverage the margin buffer is thin enough that a position can be liquidated before its stop-loss executes — losing the entire margin instead of the one unit of risk that was planned. Found via real account data: 5 of 39 trades were liquidated this way, accounting for 68% of the demo account's total loss at the time. Fixed and deployed: max leverage lowered from the exchange default to 5x.

Why it mattersResolved — closed the widest gap ever measured between designed risk and actual risk.

DoneValidation discipline

Removed self-correlated signal conditions

Three conditions treated as independent turned out to draw on the same underlying data, correlated at 99.6% — meaning the system effectively had two conditions, not three. After the fix correlation fell to 18.6%, and part of the edge we thought we had disappeared with it.

Why it mattersEliminated a phantom edge from earlier backtest results.

DoneTrading costs

Funding modelled at real settlement times

Funding is not charged continuously but at fixed times of day; a position opened and closed between two settlements pays nothing. The previous model spread it evenly across hours held and so overcharged most trades. The rate was also recalibrated against funding actually paid on the account rather than an exchange-wide average.

Why it mattersSimulated costs now track the real bill instead of an estimate.

DoneInfrastructure

Parallelised parameter fitting

Fitting one instrument fell from roughly 45–50 minutes to under 2 minutes, which is what made it possible to go from a few dozen configs to several hundred.

Why it mattersTesting scale grew nearly 8x on the same infrastructure.

DoneTrading costs

Charged trading fees and funding to every backtested trade

Backtests previously deducted no costs at all, while the live trading path recorded profit and loss net of fees — the two sides were measured with different rulers. Costs are now charged the way the exchange bills them: against position notional, converted to R via the stop distance.

Why it mattersResults are now directly comparable with real trading outcomes.

Machine-written

Deployment log

Each row is written by the system at deploy time, carrying the matching commit hash in the repository. The delta columns show what changed since the previous deployment — these figures cannot be entered by hand.

Date Version InstrumentsTrades backtested Configs with costs ReportsRoadmap done
02/09/20262e50084
Merge origin/main into gen-1 (sync after PR #54)
182100402451207/17
29/08/20267511b83
session_archive: add written summary of 2026-08-29 Windows-host session
182100402451207/17
29/08/20264f632df
Merge pull request #47 from uniconn-corp/feature/fundraising-site
182100402451205/15
29/08/202680c22e8
calibrate: prune 62 dead legacy keys shadowed by cost-aware entries
182100402451205/15
29/08/2026341c570
Merge pull request #46 from uniconn-corp/gen-1
18210040 -1839245 +70120 +865/15
28/08/2026c06f3d8
reports: rebuild BTC/ETH/AAVE with fixed calibration-key lookup
18211879175345/15
28/08/2026233be51
docs: add per-coin parameter table (glossary + actual saved config values)
1821187917534 -305/15
28/08/2026c0a363a
Merge pull request #40 from uniconn-corp/gen-1
18211879 +1857175 +53645/15
28/08/2026f0b3990
autotrade: separate margin cap for demo vs real money
182 +3110022 +2839122 +98644/14
26/08/2026e18ede6
calibration: sync from VPS after bulk-calib-top500 finished (60/60)
151718324644/14
26/08/20261ed5646
Merge pull request #31 from uniconn-corp/feature/fundraising-site-ux-ui
151718324644/14
25/08/202641b8da3
Merge pull request #30 from uniconn-corp/gen-1
151718324644/14

Showing the 12 most recent deployments of 18 total.

Support / Donate / Sponsor the Project

If you find the data or backtest results on this page useful, a voluntary USDT contribution is welcome — no return or entitlement attached.

Network Solana · USDT (SPL)
Wallet address
GXB1waW2uY58HvQsUfxUXAnq9jwef9YQV1SMoBanEMcu

⚠ Only send via the Solana network. Sending on a different network loses the funds permanently, with no way to recover them.

Entirely voluntary — not an investment, and it does not come with any return or entitlement.

Try it & give feedback on Telegram

Join the Telegram bot for real-time market-scan alerts — send /start to subscribe, /stop to unsubscribe anytime — or send feedback directly to the team.

⚠ Not seeing the QR or bot info? Ad blockers (AdBlock, uBlock Origin...) sometimes hide social-media widgets by mistake — try disabling it for this page and reload.

@fluck_pump_bot ↗

The bot sends real-time market-scan alerts. For any feedback or questions, message the bot directly.

Where visitors come from

Shaded by visit count per country — country-level only, no individual location is stored for anyone.