The promise: Every trade idea on this platform is scored using the 8 factors below. The formulas are public. The weights are public. You can verify each pick by hand. Nothing is generated by AI prose or random — every score traces to live data through documented math.
The veto rule: A pick must pass
all hard vetos AND score ≥ 48 / 80 total. Otherwise the desk shows "no qualifying setup today." Better to skip than chase.
Requires live data: Scoring is disabled when the site is in DEMO MODE. The brain refuses to publish a TOTD pick from synthetic prices.
Check data status.
🎯 Tier system
A
64-80
High-conviction. Full size. Multiple factors aligned. The setup the desk wants.
B
56-63
Solid. Take with discipline. One factor weak but everything else aligned.
C
48-55
Watch / half-size. Multiple borderline factors. Paper-trade for learning.
F
< 48
Skip. No TOTD pick published. Sit out.
🚫 Hard vetos (any one = skip the symbol)
Liquidity floor: Average daily volume < 1M shares. Spreads are too wide, slippage will eat your edge.
Earnings tomorrow: The brain can't predict earnings outcomes and IV crush will distort everything.
Halted / pre-market only: No primary-session liquidity = no setup.
Penny stocks: Last price < $5 — too easy to get manipulated.
Wild gap: Today's open is >5% from yesterday's close — too volatile, the setup is broken.
No live data: Site is in DEMO MODE — no pick is generated.
📊 The 8 factors (10 pts each, 80 total)
Today's volume vs the 20-day average. Sweet spot is 1.2-3.0x average — enough conviction to mean something, not so chaotic that it's a panic.
101.5x to 3x average volume
71.0x to 1.5x average
40.7x to 1.0x
0< 0.7x (no conviction) or > 5x (panic/news)
Formula: vol_today / vol_20d_avg
2. Trend strength
0 - 10 pts
Price relative to its 20-day and 50-day moving averages. A clean trend (above or below both, slopes aligned) earns full points. Conflicted (between them) gets partial.
10Price above 20MA & 20MA above 50MA — both risingor full inverse for shorts
7Price above 20MA, but MAs flat
4Price between MAs (conflicted)
0Choppy / no clear trend on either MA
Formula: aligns price vs MA20, MA20 vs MA50, slope of each
Today's % change combined with 5-day return. Reward decisive moves in a clear direction; penalize chop.
10Today ±2% AND 5d return ±5% same direction
7Today ±1% AND 5d in same direction
4Today within ±0.5% (drifting)
0Today opposite direction of 5d (reversing)
Formula: sign(today_pct) × sign(5d_return) × magnitude_score
4. Volatility quality
0 - 10 pts
ATR% (average true range / price). Too quiet = no movement to capture. Too wild = stops get hit by noise. Sweet spot: 1-4%.
10ATR% between 1.5% and 3.5% (active but tradeable)
71.0-1.5% or 3.5-5%
40.5-1% (too quiet) or 5-7%
0< 0.5% (dead) or > 7% (chaos)
Formula: (14-day ATR / current price) × 100
5. Setup clarity
0 - 10 pts
Distance from a meaningful level — recent high/low, key MA, round number. Trades fire best from clear levels because stops are obvious.
10Within 1% of a level AND just bounced or broke
7Within 2% of a major level
4In the middle of a range — no clear level nearby
0Trapped between two opposing levels
Levels considered: 20-day high/low, 50MA, 200MA, round numbers, prior swing pivots
6. Risk/reward
0 - 10 pts
Distance from current price to the logical stop vs the logical target. The math has to work for the strategy to pay out long-term.
10R:R ≥ 2.5:1
7R:R 2.0 to 2.5
4R:R 1.5 to 2.0
0R:R < 1.5 — math doesn't work
Stop = nearest level beyond entry. Target = next major level OR 1.5×ATR.
7. Catalyst neutrality
0 - 10 pts
No major catalyst in the trading window. Pure technical setups score high; pending earnings/Fed/news scores low because the model can't predict event outcomes.
10No earnings within 5 trading days, no FOMC, no scheduled news
7Earnings 5-10 days out
4FOMC or earnings within 2-5 days
0Earnings tomorrow, FOMC today — auto-veto
When no earnings feed is connected, score defaults to 7 (neutral).
The brain's learned probability of win for this exact feature pattern. Only counts when the model has been trained on ≥20 real (non-mock) labeled outcomes. Until then this factor defaults to 5 (neutral) so the methodology works from day one without contaminated learning.
10Model probability ≥ 70%(requires 50+ labeled live trades)
760-70%
450-60%
0< 50% (model thinks losing setup)
Until the brain is calibrated on real data: defaults to 5 (neutral). See
Data Truth.
Most retail trade-idea services hand you a prediction without showing the math. That's a black box and you can't sanity-check it. With the methodology above:
• Every score traces to a live data point (volume, ATR, price vs MA, etc).
• Every weight is fixed and public — no hidden tuning per pick.
• The methodology fails loud — if it can't score, it shows "no pick."
• The ML overlay is only one factor out of 8 and only activates once the brain has real training data. The platform works without ML from day one.
• You can compute it on a napkin if you don't trust the code.
The point isn't that this methodology is the only good one. The point is it's explicit and falsifiable. When a pick goes wrong, you can audit which factor was overweighted.
Trade of the Day uses this exact methodology against the live data feed. Every page load re-runs the scoring on every symbol in the universe and publishes the top-scoring symbol — or "no pick" if nothing crosses the 48-point threshold.