Exposure
Lot sizes, leverage and aggregate risk relative to capital.
See an example check
Several small positions may form one large exposure if they depend on the same market.
THE FINBASE METHOD
A strategy is selected. Then put to the test.
A good-looking chart is not enough. We seek to understand how a strategy wins, how it loses and what could put it under pressure. This is the selection process described by FINBASE.
Checks work together ↓Time is part of the review
12months minimumof live account history
+3–6additional monthsof observation in live conditions
=15–18months of historyminimum, depending on the observation period
With 12 months of initial history and 3 months of observation, the minimum is 15 months. With 6 months of observation, it becomes 18 months. This schedule is neither automatic approval nor a guaranteed launch date.
01
Research starts before the numbers
FINBASE works with many algorithm developers and traders to identify interesting approaches. Candidates also contact the team directly to join the community. A referral or application is not enough for approval: the approach, operating conditions and track record must be reviewed.
02
First, a live track record
The process requires at least one year of history on a live trading account before the in-depth review. A screenshot or backtest does not replace this record. The review examines data continuity, trades, deposits and withdrawals, and the conditions behind the results. A balance increase caused by a deposit must not be mistaken for performance.
03
Look at the risk behind the result
Algorithms, analysis bots and AI assist the review of trading histories. They help identify risk concentrations, inconsistencies and changes in behaviour. Traders interpret the alerts: AI does not approve a strategy on its own and is not evidence of reliability.
Concurrent positions
Number of open positions and overlapping risks.
See an example check
Opening several positions in gold does not necessarily provide diversification.
Duration & frequency
Holding times, frequency and behaviour around economic releases.
See an example check
An unusual increase in how long losing positions are held may reveal a behavioural change to review.
Actual costs
Spreads, commissions, financing and actual execution prices.
See an example check
A strategy profitable before costs may lose its edge when execution costs are included.
04
Do not confuse balance with actual risk
Drawdown measures the fall from a previous peak. Analysis must consider equity, including gains and losses on positions still open, not only the balance after trades close. An account may look stable while carrying a substantial unrealised loss. Without sufficiently detailed equity data, actual intratrade drawdown cannot be reconstructed reliably from closed trades alone.
Balance: closed tradesEquity: including open positions
05
Test what the bot has not already learned
Overfitting occurs when settings match a particular history too closely without being robust elsewhere. Stress tests expose the algorithm to new or worsened simulated conditions: synthetic price paths, periods not used for parameter selection and different execution conditions. The aim is to identify excessive dependence on specific years or conditions. A simulation never perfectly reproduces live markets and cannot prove the absence of overfitting.
History used for parameter selectionScenarios not used for parameter selection
A different market regime
Trends, sideways markets and sudden moves: check that behaviour does not depend on a single historical pattern.
Worse execution
Widen simulated spreads and slippage to observe their effect on entries, exits and costs.
Less perfect settings
Vary parameters around the selected settings: excessive fragility should be reviewed, not hidden by optimisation.
06
Then, 3 to 6 months in live markets
After the checks, the strategy is observed live, generally for three to six months. Traders monitor its behaviour and compare it with what was described and analysed. Alerts cover changes in risk, concurrent positions, unrealised losses, costs and execution anomalies. This phase is separate from simulated stress tests.
Observation → alert → review → decisionEducational diagram · no performance data shown
07
An alert must lead to action
A confirmed critical warning removes the strategy from the selection process. Having no alerts is not enough on its own: checks and observation must be completed before a publication decision. Review thresholds reflect each strategy’s behaviour and risk, rather than a single value applied to every system. A material change to the system or its settings calls for a new review of the affected scope.
Confirmed critical flag: excluded from the process
Checks passed: publication decision
08
What this method helps you understand
This page describes FINBASE’s internal selection method. It is not an individual report for each strategy, a regulatory certification or an independent audit. The stated durations describe the admission process; they do not replace the dates and evidence for each individual review. Check the relevant strategy’s history and conditions, then the records available at the broker. Even a selected strategy can incur significant losses.