How the Risk Score Works
Every asset on ZonedInOut gets a score from 30 to 98. The score answers one question: is now a historically favorable time to enter? It's not a prediction. It's a risk measurement — built from five independent factors that combine to say how extended or how pulled-back an asset is, relative to its own recent history and trend structure.
The trade that started too many portfolios the wrong way
November 2021. NVIDIA was trading around $329 on a split-adjusted basis. It had climbed steadily for three straight months. Every financial media outlet was running some version of the same headline: the AI future is here, the chip shortage is a tailwind, NVDA is unstoppable. The momentum was undeniable. A first-time investor — let's say they'd been watching it go up for weeks, reading the coverage, building conviction — finally buys.
Three months later: $230. Six months later: $180. By October 2022, NVIDIA was trading near $108 — a 67% decline from the peak, and a 45% loss from that November entry. The investor who bought wasn't uninformed. The company was genuinely good. The chip shortage was a real tailwind. The AI story played out exactly as predicted. NVIDIA eventually became one of the greatest performing stocks of the decade.
The entry timing was just terrible. The investor bought after three months of straight gains, at a point where the stock was trading roughly 30% above its 20-day moving average. The math was screaming extended. No one was looking at the math.
Why it keeps happening
The brain is wired to treat recent price history as evidence of future direction. When something has gone up for three months straight, it feels like proof it will keep going up. Psychologists call this recency bias — the tendency to weight recent events far too heavily when predicting the future. In markets, it's compounded by survivorship: the people you hear from are the ones who bought NVIDIA in 2019 and held. You don't hear from the ones who bought in November 2021, because they went quiet.
FOMO does the rest. When a stock has been climbing for weeks, every day you don't own it feels like money you're leaving on the table. The discomfort of watching it go up without you is, for most people, psychologically worse than the actual loss of being wrong. So they buy — at exactly the moment the asset is most extended from its historical average.
The market is brutally good at making the worst entry look like the most obvious buy. Peak prices come with peak headlines, peak analyst upgrades, and peak social media conviction. That's not coincidence. That's how tops form.
The mechanism behind mean reversion
Price always deviates from its average, and then reverts. This isn't mystical — it's how competitive markets work. When a stock runs 25% above its 20-day average in three months, something mechanical happens: early holders who bought at lower prices start taking profit. Institutional funds running momentum strategies begin to rebalance. The next marginal buyer has to enter at a higher price than the previous one, which makes the position less attractive, which slows inflows, which reduces the price pressure holding it up. The rubber band snaps back.
The 20-day moving average is the anchor — the average price at which the last 20 trading sessions of buyers and sellers agreed to transact. It represents something like the market's short-term sense of fair value. When current price diverges significantly from that anchor, it quantifies the degree of disagreement between where the stock is trading and where it has been trading.
Distance from the anchor is risk. The further above it you buy, the more likely you are buying the reversion, not the continuation.
The model — five layers
The score is built in five independent layers. Each adds precision to the raw signal. Higher score = lower entry risk. When NVIDIA was at $329 in November 2021, the primary layer was screaming extended. A tool running this model would have flagged it clearly. The math was there. No one was running it.
How far the current price sits above or below the 20-day MA, measured in ATR (Average True Range) units. Using ATR instead of raw percentage normalizes for volatility — a 5% gap means different things for a calm ETF vs a high-beta growth stock. Negative ATR distance (price below MA20) → highest scores. Positive ATR distance (price extended above MA20) → lowest scores.
Price above a rising MA200 adds up to +6 pts — the pullback you're buying has the long-term current behind it. Price below a falling MA200 subtracts up to −6 pts — mean reversion in a downtrend is less reliable. Also drives the regime classification that adjusts signal labels (see below).
RSI below 35: +3 pts (oversold, adds conviction to a high base score). RSI below 25: +5 pts (deeply oversold). RSI above 65: −3 pts. RSI above 75: −5 pts (overbought, caution even if price looks pulled back).
A down day on low volume (<75% of 20-day average): +2 pts — low-conviction selling, suggesting the dip is likely temporary. A down day on high volume (>175% of average): −2 pts — institutional selling, the move carries weight.
When Bollinger Bands are at their tightest range in months (a squeeze), a significant directional move is usually imminent. A high score during a BB squeeze with a rising MACD histogram: +5 pts — compressed spring with upward pressure. Squeezed with MACD falling below zero: −3 pts.
The five layers combine into a raw score (clamped 30–98), then a trend regime adjustment applies: assets in a confirmed uptrend receive a score floor so a healthy pullback within a bull trend is never mislabeled as "Avoid." The tiers break down like this:
Signal tiers
| SIGNAL | SCORE | MEANING |
|---|---|---|
| STRONG BUY | 90–98 | Deep pullback — near or below MA20. Lowest entry risk in the model. |
| BUY | 79–89 | Moderate pullback — favorable entry window, mild extension. |
| ACCUMULATE | 71–78 | Slight extension — acceptable to build gradually, not ideal for a full position. |
| HOLD | 64–70 | Extended — don't add here, hold existing positions and wait. |
| EXTENDED | 50–63 | Significant extension — elevated entry risk, probability of near-term mean reversion is higher. |
| AVOID | < 50 | Heavily extended or in a structural downtrend. Statistically poor entry timing. |
| RIDE TREND | any | Regime signal only — asset is in a confirmed strong uptrend (price >20% above a rising MA200). Pullbacks within a bull trend are normal; fighting the trend is statistically worse than holding. |
The trend regime — why the same score means different things
The trickiest problem in mean-reversion scoring: sometimes "Avoid" is wrong. A stock that's been in a confirmed uptrend for a year, sitting 25% above its MA200 with the MA200 still rising — calling every dip "Avoid" is backwards. The backtest data proves it: on assets like NVDA during its 2023–2024 run, "Avoid" signals posted 83%+ win rates because every dip was a buying opportunity within the trend. The model was technically correct about extension but giving the wrong action label.
The regime layer fixes this. The model classifies each asset's current market position into one of three regimes:
- strong_trend — price more than 20% above a rising MA200. Score floor: 55. Pullback bonus: +5 pts if price dips below MA20. Signal: Ride Trend instead of Avoid.
- trend — price more than 8% above MA200, not falling. Score floor: 47. Pullback bonus: +3 pts. Signal: Extended instead of Avoid.
- neutral — everything else. No adjustment. Standard labels apply.
These adjustments don't change the raw score dramatically — they prevent it from misfiring at the extremes. A score of 85 in a strong uptrend with RSI oversold and low-volume selling is a materially different setup than an 85 in a flat market with no trend context. The model now knows the difference.
Getting out: the exit plan
Everything above is about entry. It's incomplete without an exit plan, so every asset's setup panel also computes a stop-loss and two profit targets — not as a flat percentage, but from that asset's own recent volatility (ATR, the same measure Layer 1 uses to normalize the entry signal). A stop that's 2.0× ATR below the entry means something different for a calm large-cap than a volatile small-cap, which a fixed "-8%" rule never accounts for.
The math: stop = 20-day average − 2.0× ATR. Risk = the distance from entry to that stop. Target 1 sits at 1.5× that risk above entry (take partial profit here). Target 2 sits at 2.5× the risk (let the rest run).
We tested this the same way the entry signal gets tested: backtested, not assumed. 1,991 historical Buy Zone entries (score 79+, the same threshold used everywhere on the site) across 5 years and the same 23-name universe the edge report covers, walked forward day by day against real intraday highs and lows to see which level actually got hit first.
| STOP WIDTH | STOP HITS FIRST | TARGET 1 FIRST | T1→T2 FOLLOW-THROUGH |
|---|---|---|---|
| 1.0× ATR | 74.7% | 25.3% | 78.2% |
| 2.0× ATR — LIVE | 67.1% | 32.0% | 77.6% |
| 2.5× ATR | 62.7% | 34.5% | 71.2% |
| 3.0× ATR | 54.8% | 38.3% | 62.3% |
Two honest findings came out of this, and neither is flattering enough to have been worth guessing at instead of testing:
The stop is the single most likely outcome at every width we tested — even a stop three times wider than what we ship still gets hit before either target more often than not (54.8%). Widening the stop helps, but there's no width in a reasonable range that flips those odds. We chose 2.0× ATR because it's the best tradeoff we found: meaningfully fewer premature stop-outs than the tighter setting, without giving up much of the 1→2 follow-through quality. Wider than that, the numbers get worse in a different way — trades take much longer to resolve, and a growing share never resolve at all within a reasonable window.
A stop-out is not proof the entry was wrong. The entry signal's actual edge — beating a plain SPY buy-and-hold — is measured separately in the edge report, over 5/20/60-day raw returns with no stop attached at all. A tight mechanical stop and a mean-reversion setup that can take one to three weeks to play out are pulling in different directions. If you use these levels, treat the stop as real risk control, not as a verdict on whether the score was right.
What it doesn't tell you
The score won't tell you if the company is good. It won't tell you if next quarter's earnings will beat. It won't call the bottom with precision. What it quantifies is entry risk relative to recent history — not future returns.
A score of 94 in a stock with declining fundamentals is still a score of 94 in a bad company. The timing might be right. The investment might still be wrong. This is a timing filter, not a stock picker. Use it to answer "is now a good time?" — then apply your own research to answer "is this a good investment?"
The score has zero awareness of why a stock is falling. The model reads price, volume and RSI — nothing else. If a company crashes because a product got recalled, a lawsuit landed, or fraud came out, the score doesn't know any of that happened. All it sees is price stretched far below its 20-day average, and a stretch like that reads exactly the same whether it's a healthy panic dip or the start of a real collapse. Pushed far enough, that crash can price out as a 90+ "Strong Buy" — the score cannot, on its own, tell the difference between a stock that's cheap and a stock that's broken. The KNIFE flag on the rankings table exists specifically to catch this pattern (a 79+ score down more than 10% over three months) — historically that cohort hit its stop 81.6% of the time before it hit a target, per our own backtest of this exact failure mode. That flag catches the symptom. It still can't read the news, so it can't tell you the cause — that part is still on you.
The Opportunities Board
The Opportunities Board goes further than a single signal. It composites five data points into a conviction score that surfaces setups where multiple independent factors are aligned — not just one indicator ticking a box:
- ZIO Score (0–35 pts) — the base entry risk signal, weighted as the primary input
- Analyst price target upside % (0–30 pts) — Wall Street consensus via Yahoo Finance; how much room does the street think is left?
- RSI oversold level (0–20 pts) — RSI below 40 adds momentum confirmation; the selling pressure is real, not theoretical
- Score momentum (0–15 pts) — a score moving in the right direction is worth more than one that has been sitting still
- Buy-the-dip bonus (0–10 pts) — high score combined with a negative 24-hour price change: the asset is pulling back while the signal strengthens
Data sources
- Yahoo Finance — the primary data layer: live prices, RSI, 20-day and 200-day moving averages, analyst price targets, insider transaction filings, and institutional ownership — all from a single reliable feed updated in real time
- ForexFactory — the economic calendar that professional traders actually use; impact ratings (high / medium) for every scheduled event, so you know when a CPI print or FOMC decision could override the technical signal
- Alternative.me — the Crypto Fear & Greed Index, which aggregates volatility, market momentum, social sentiment, and dominance into a single number that has historically flagged crypto extremes with surprising accuracy
- SEC Form 4 — the mandatory insider trading disclosure that executives and directors must file within two business days of a transaction; sourced via Yahoo Finance, it's the closest thing to watching where the people who know the most are putting their own money
Disclaimer
ZonedInOut provides quantitative data and analysis for informational purposes only. Nothing on this site constitutes financial advice, investment advice, or a recommendation to buy or sell any security. Past signal performance does not guarantee future results. Always do your own research and consult a qualified financial advisor before making investment decisions.