Current stage
User risk tags and User Performance Score are currently in the model-training, rule-validation, and continuous-optimization stage.
The current version combines defined trading indicators, thresholds, and model analysis results. Tag names, trigger conditions, thresholds, score logic, risk levels, analysis dimensions, and display methods may change as the model and product evolve.
At this stage, the feature is provided as a free add-on capability of Finger Manager and is not charged separately.
AI and model usage
The goal of user risk tags is not to let AI replace institutional risk personnel. The system first extracts quantifiable behavioral features from trading data, then generates labels according to existing rules and model analysis.
For example, a high recent order count may generate a high-frequency tag; a high single-product ratio may generate a product-focus tag; short average holding time may generate a short-term trading tag; and a high Buy ratio may generate a Buy-biased tag.
The design principle is: data analysis, system-generated tags, visible evidence, and human review. AI does not make the final risk decision automatically.
Important disclaimer
User risk tags, User Performance Score, risk levels, and related analysis results are in training and continuous optimization. Results may contain bias, misjudgment, omissions, delay, or inconsistency with actual circumstances.
Finger Manager does not guarantee that tags, scores, or risk signals are complete, accurate, or suitable for any specific business purpose. They are only auxiliary references for customer portrait, operations analysis, trading behavior analysis, and risk observation.
They do not constitute formal risk rating, credit rating, compliance conclusion, investment advice, trading advice, suitability judgment, forecast of future profit or loss, or automatic account-processing decision.
Institutions should not rely only on a single tag, score, or risk level to freeze accounts, restrict trading, reject service, terminate relationships, or take other actions that may affect customer rights. Practical decisions should combine original order data, current positions, account information, institutional risk rules, and necessary manual review.
Win-rate tags
The system generates win-rate tags according to the percentage of profitable orders among valid orders. Win rate only reflects the count ratio of profitable orders; high win rate does not necessarily mean high profit quality.
Elite winner Win rate >= 70%. Very high win rate in the current sample.
Strong winner 60% <= win rate < 70%. Strong win-rate performance.
High win rate 50% <= win rate < 60%. Win rate is relatively high.
Low win rate 40% <= win rate < 50%. Win rate is relatively low.
Weak performer 30% <= win rate < 40%. Current trading performance is weak.
Poor win rate Win rate < 30%. Low win rate in the current sample.
Payoff-ratio tags
Payoff ratio observes the structural relationship between average profit and average loss. The system combines win rate and payoff ratio so that it does not evaluate performance by profitable-order count alone.
Strong return Payoff ratio >= 300%. Single-order profit is strong relative to loss.
Healthy return 150% <= payoff ratio < 300%. Overall P/L structure is healthy.
Balanced return 80% <= payoff ratio < 150%. Profit and loss are relatively balanced.
Weak return 50% <= payoff ratio < 80%. Profit quality is relatively weak.
Loss pressure Payoff ratio < 50%. Single-order loss has a strong impact on overall result.
Average holding tags
The system uses average holding duration to identify the customer's primary trading cycle. Holding duration describes trading style; it does not by itself mean high or low risk.
Seconds-level trading Average holding < 1 minute. Very short-cycle trading.
Quick in and out 1-10 minutes. Short-term trading.
Intraday short-term 10-60 minutes. Intraday short-cycle trading.
Intraday holding 1-8 hours. Mainly same-day holding.
Overnight tendency 8-24 hours. Shows an overnight tendency.
Long holding user More than 24 hours. Long average holding cycle.
Recent trading activity
The system uses the number of orders in the last 7 days to identify recent activity. This tag is time-sensitive; a historically active customer may still be silent if no orders were placed recently.
High-frequency active >= 50 orders in the last 7 days.
Active trading 20-49 orders in the last 7 days.
Stable trading 5-19 orders in the last 7 days.
Low-frequency trading 1-4 orders in the last 7 days.
Silent user 0 orders in the last 7 days.
Instrument and market preference
The system analyzes Symbol and market-category ratios to identify product preference and concentration. These tags describe what the customer tends to trade and how concentrated the behavior is.
{Symbol} focus A single Symbol ratio >= 80%.
Gold / US stock focus Generated when the focused symbol belongs to gold or stock categories.
Crypto / Forex / Gold / Stock preference Generated when the corresponding market ratio reaches 70% or above.
Multi-market trading Generated when two or more markets contain actual trading records.
Buy/Sell direction and directional win rate
The system analyzes Buy and Sell order ratios to determine whether the customer has a clear directional preference. It can also compare Buy and Sell win rates if both sides have enough samples.
Buy-biased trading Buy ratio >= 70%.
Sell-biased trading Sell ratio >= 70%.
Balanced direction Buy/Sell ratio gap <= 10%.
Buy advantage Buy win rate is at least 20 percentage points higher than Sell win rate.
Sell advantage Sell win rate is at least 20 percentage points higher than Buy win rate.
Minimum sample requirement Both Buy and Sell need at least 3 valid orders before directional win-rate tags can be generated.
User Performance Score
In addition to individual tags, Finger Manager generates a User Performance Score from multiple dimensions. The score is an internal trading-performance analysis indicator, not a credit score or formal risk rating.
Score structure Win-rate score 25, payoff quality 30, profitability 20, trading behavior 15, risk penalty up to 10.
S / A / B S >= 90, A >= 80, B >= 70. Used for stronger or more stable performance ranges.
C / D / E C >= 60, D >= 40, E < 40. Used for ordinary active users, risk-observation users, and high-risk users.
Example Performance Score 78 / B means the current data places the customer in the B range.
Risk signal tags
Besides behavior tags, the system can generate risk-signal tags that should be reviewed further by authorized staff.
Small profit, large loss Usually appears when payoff ratio is about 50%-80%; the customer may make many small profits but suffer relatively larger losses.
Heavy loss tendency May appear when payoff ratio is below 50%, or average loss is about two times greater than average profit.
Light loss May appear when customer ROI is around -10% to -3%.
Loss pressure May appear when ROI is below about -10% or overall trading results show clear losses.
Insufficient sample Appears when there are fewer than 5 valid trades; other tags and score should be treated with lower confidence.
Long-holding risk May appear when average holding duration is clearly long, indicating extended market exposure.
Concentrated instrument risk May appear when one instrument reaches 80% or more and its win rate or payoff quality is insufficient.
High win, low quality May appear when win rate is 60% or above but payoff ratio is below 80%.
Recent losses May appear after about 3-4 consecutive losing orders.
Consecutive losses May appear after 5 or more consecutive losing orders.
Floating loss pressure May appear when current unrealized loss is greater than realized profit.
Long-holding loss May appear when average holding duration is more than 24 hours and total trading P/L is negative.
Evidence and manual review
Finger Manager tries to display the main evidence behind each generated tag. For example: 56 orders in the last 7 days generates high-frequency active; XAU/USD accounting for 82% generates XAU/USD focus; average holding time of 04m 18s generates quick in and out; Buy ratio of 74% generates Buy-biased trading.
This lets back-office users understand not only which tag was generated, but also why it appeared.
User risk tags are not permanent customer attributes. As new orders are created, win rate, payoff ratio, trading frequency, holding duration, product preference, direction, profitability, and risk signals may all change. The system recalculates the portrait based on the latest data.
Usage note
The overall logic is: customer trading data -> behavior feature calculation -> tags and Performance Score -> risk signal identification -> evidence display -> back-office review.
The goal is not to classify customers simply as good or bad. It is to help institutions turn large trading datasets into customer behavior portraits that can be quickly read and further analyzed.
During the testing stage, user risk tags and Performance Score remain in model training and continuous optimization, and are provided as a free add-on capability of Finger Manager. Outputs may contain bias or misjudgment, and should only be used as references for operations analysis and risk observation. They should not be used as the sole basis for formal risk determination or customer account handling.
