Weinstein Live — Stage Analysis

Click any row in the table below to open its annotated chart. Each one is labelled the way Weinstein teaches it:

• Shaded bands, named — every stage the stock has passed through: STAGE 1 base STAGE 2 advance STAGE 3 top STAGE 4 decline. Runs shorter than 8 weeks are absorbed into their neighbours, because a stage is a phase of months, not a weekly flip.
• The red line is the 30-week MA — the single most important object on the chart. “Stocks trading beneath their 30-week MAs should never be considered for purchase, especially if the MA is declining.”
• The green dashed line is the ceiling of the most recent base, with an arrow at the point price broke through it — where Stage 2 begins.
• The purple dotted line is the protective sell-stop, placed below the correction low or the MA (whichever is lower) and tucked just under a round number, with the risk from today's price. “Ask eight out of ten investors who bought stock XYZ at 40 how much they've lost now that the stock is down to 20, and they'll answer nothing!… That's ridiculous! Your stock is only worth the price of the last print on the tape.”
• Volume panel — the dashed red line is 2× the 4-week average, Weinstein's breakout threshold; bars that clear it are red. “Never trust a breakout that isn't accompanied by a significant increase in volume.”
• A written read-out accompanies every chart — how Weinstein would read it in words, followed by the specific price levels at which the stage would change. Naming those levels is his instruction: “write down the price that each would need to break out.”
• RS panel — Mansfield relative strength: the stock-divided-by-Nifty line measured against its own 52-week average, so the zero line is real. Above zero the stock is beating its own year; below it, it is not — and “absolutely never buy a stock… if the relative strength is in negative territory.” The level is enforced; the direction is shown but not enforced, because whether a line is “in a downtrend” changes with the window you measure over.
• Entry — the green dashed line with a tick above it. For a stock still in Stage 1 this is the base ceiling plus a tick, and the order is a resting GTC buy-stop that fires only on the breakout. For one that has just crossed, it is today's price. Where a half-and-half applies, the second half is drawn in the same green, finely dotted, at the pullback level — and only over the right-hand part of the chart, since it is a level for now and not for history. “Buy half of your intended position on the initial breakout, and the other half if the stock pulls back close to the breakout price.”
• Exit — the purple dotted line, and it is calculated before the entry, not after. “Never enter your order to buy until after you’ve calculated exactly where your protective stop should be placed.” The order ticket states the rupee distance and the percentage from the price you would actually pay, and refuses the trade above 15%.
• Trend change — the level is named in figures, not described. The 30-week MA is the trend-change price: a close through it, with the average turning, changes the stage. Every chart carries it as a labelled level at the right edge, and every read-out repeats it in words, because writing the level down in advance is the instruction — “write down the price that each would need to break out.”
• Uptrend line — hidden by default, one button to show it. Drawn only where the advance has made three or more rising correction lows (he labels three: “points D, E, and finally F”) and fitted by least squares, so it is a computed line rather than one drawn by eye. 19 of the 59 names qualify. Treat a break as a signal to tighten the stop, not as the exit itself — the stop is what takes you out.
• Sizing is deliberately NOT on the chart. How much to commit is a portfolio decision, not a property of one stock's chart, so it lives under Position sizing and trade management instead.

Weinstein spends more of the book on what happens after the entry than on finding it. The entry is one decision; the size, the stop and the trail are the ones that decide the outcome. None of this is a property of a single chart, which is why it was taken off them — it is set once, for the portfolio, here.

1 · Equal rupee amounts, and enough of them that one loss cannot matter

He does not size by conviction. Every position is the same rupee amount — “if we were investing $100,000 in 10 stocks, each position should be approximately $10,000” — and the number of positions is chosen so that a stop being hit is an event you barely feel:

2 · The stop is calculated before the order, not after the fill

“Never enter your order to buy until after you’ve calculated exactly where your protective stop should be placed.” The sequence matters because it is the only sequence in which the 15% cap can actually reject a trade. Work it the other way — buy, then look for somewhere to put a stop — and the stop gets placed wherever the loss is tolerable rather than where the chart says, which is how a 15% rule becomes a 40% loss.

3 · Half and half — he never commits a whole position at one price

“Buy half of your intended position on the initial breakout, and the other half if the stock pulls back close to the breakout price.” The second half is not a doubling-down; it is the same position bought at two prices, and it comes with a test attached. On the pullback you are watching whether volume contracts on the way back — a pullback on heavy volume is a failure, not a discount.

The window closes, and the page says so when it has. Once the trailing stop has risen above the old breakout price, a pullback to that breakout would trigger the exit rather than an entry — so the trade becomes a single entry, take it or leave it. And where price has already slipped back below the level it broke out through, that is his red flag: “did XYZ just nudge above the critical resistance level and then fall back below it, which always signals a red flag warning.”

4 · Trailing the stop — and the specific mistake of doing it too soon

The stop only ever moves up. But not on every new high, which is the error that shakes people out of good positions:

On a half-and-half position he runs two stops, not one — “one half of your position is now protected with a stop at point C, and one half is set very tightly at rising points D, E, and finally F.” The loose half rides the 30-week average; the tight half rides the uptrend line through the correction lows. That is what the uptrend-line toggle on each chart is for: it is the tight stop's track, not decoration, and it is drawn only where three or more correction lows exist to fit it to.

When the 30-week average flattens, the tactics change. A Stage 3 top forms while price is still making highs, so the rule inverts: move the stop under the last correction low even if that sits above the average. “The probabilities are high that a Stage 3 top is starting to form.” The order ticket on each chart switches to this wording on its own when the slope goes flat.

5 · Cash is a position, and the portfolio has a stop too

“Don’t feel you have to be 100 percent invested at all times.” He puts it in sporting terms: “many athletes perform incredibly well in warm-ups, only to choke once the actual game begins… use only a portion of your capital” until the method has actually made money for you. The amber box under the market gate at the top of this page computes what that means today from the current gate reading and your own position count.

And above the individual stops sits one on the whole account: “I strongly suggest that each weekend you tabulate the value of your portfolio… if you see your assets drop by 20 to 25 percent, you should immediately move to the sidelines.” Weekly, on the weekend, same as the scan — which is the cadence this whole screen is built to.

Give it a name or symbol and it runs the whole screen on that stock the way Weinstein would work through a new idea — market first, then the stock's own stage, then each filter with the sentence from the book behind it, then the risk — and ends with a verdict.

cd "Investor Frameworks/WEINSTEIN" && ../../.venv-charts/bin/python tools/lookup.py RELIANCE

Accepts a symbol or a name — lookup.py INFY, lookup.py "tata motors". Add --json for machine-readable output. It resolves the company on Screener and echoes back the name it actually found before analysing, so a wrong match is visible rather than silent; where a query is ambiguous it lists the other candidates.

Why this is a command and not a box on this page: the lookup has to fetch live weekly data, and Screener sends no Access-Control-Allow-Origin header — a browser cannot call it from this domain. Doing it here would need a serverless function, which means changing the Vercel config that serves the main dashboard. That is not a change worth risking untested, so it is a follow-up rather than a shortcut taken now.

Every rule above is mechanical, and none of it survives contact with the person applying it. Weinstein is unusually direct about this: the method fails in predictable, human ways, and he names them.

Losing well is a skill, and it is the one being taught

“One of the real keys to winning is learning how to lose.” Not tolerating losses — taking them small, on purpose, on a level written down in advance. Everything in the risk section exists so that being wrong is cheap enough to be survivable and frequent enough to be unremarkable.

The corollary is the one people refuse: the stop that takes you out a few points before a stock turns around has still done its job. “Look at those few points you gave away as the cost of insurance.” You cannot price insurance by looking only at the fires that did not happen.

The named cause is ego, not analysis

Asked why investors hold losers and sell winners, he does not reach for a market explanation. “What drives us to this destructive behavior? In one word: ego!” Selling a loser is admitting the purchase was wrong; holding it keeps the verdict pending. The market does not care either way, but the portfolio keeps the score:

Start small, because the warm-up is not the game

“Many athletes perform incredibly well in warm-ups, only to choke once the actual game begins…” — so commit only part of the capital until the method has been run, with real money, through a losing sequence as well as a winning one. Paper results are warm-ups.

Why this screen is a script and not a judgement call

Every criterion was frozen in writing before it was run, the six amendments since are logged with dates, reasons and whether any verdict moved, and the engine — not the page — decides each order, so what is displayed and what is validated cannot drift apart. That is not tidiness. It is the only defence against the ego he names: a rule that can be reinterpreted after you see the result is not a rule, and a stop you can talk yourself out of on the day is not a stop.

The screen is a script, not a snapshot. He asks for it weekly, and names the number that should make you stop: “I strongly suggest that each weekend you tabulate the value of your portfolio. Don't worry if it swings a few percent from week to week. That's the way of the market. But if you see your assets drop by 20 to 25 percent, something is wrong.” To refresh:

cd "Investor Frameworks/WEINSTEIN" && ../../.venv-charts/bin/python tools/refresh.py

• Re-fetches weekly price and volume for all 67 names plus the Nifty, with back-off on rate limits and a fall back to the cached file if a fetch fails — a failed fetch never silently becomes an abstention.
• Re-stages everything on the frozen criteria, re-computes stops and RS ranks.
• Diffs against the previous run and writes changes.json: stage transitions, verdict changes, stop breaches, and volume expansions through the 2× threshold.
• Appends a row to history.json so the distribution can be tracked over time.
• --no-fetch recomputes from cache without touching the network.

Stop breaches fire on the transition, not the state — a stock that is already below its stop is not re-reported every run. That bug was caught by running the engine twice and demanding zero changes the second time; it is now a standing check.

• The market gate rests on more than one gauge but not on all of them. Index stage analysis, breadth (% above the 30-week MA), the cumulative A–D line, the momentum index and world markets are built. Two weaknesses were named here and have now been repaired (12-Sep-2026). The A–D line no longer comes from a rolling 751-session query whose window ended near the line's own peak — it is rebuilt from NSE's own daily bhavcopy, one file per session, keyed by date, so the series is exact, reproducible, and long enough that the peak is observed rather than assumed; the liquidity filter is now applied to each day's own prices instead of today's. And world markets are no longer proxies at all: the gauge now stages eight real indices in their own currency — the Dow, S&P 500 and Nasdaq Composite, then the FTSE 100, DAX, CAC 40, Nikkei 225 and Hang Seng — with each US average voting on its own, since his standard is that “all of the market averages” agree. No FX correction is applied or needed. The ETF and feeder-fund proxies it replaced were not merely imprecise: the Hang Seng ETF in rupees read Stage 4 where the index in HKD reads Stage 3, and a Europe feeder fund read Stage 2 where the DAX and CAC 40 both read Stage 3. What is still true: the series come from FT's chart API (addressed by an id scraped from each index's own tearsheet and accepted only when the page names the index expected) and Nasdaq's own API — not from the exchanges themselves. FT's S&P 500 was checked against CBOE's official SPX history: 1,305 overlapping dates, none differing by more than 0.1%. Stooq, Yahoo, FRED, WSJ, MarketWatch and Nikkei's own site are all unreachable or bot-blocked from this machine, and Screener lists the real indices but serves an empty series for every one.
• Group stages use NSE's own indices — 81 of the 85 broad-market, sectoral and thematic indices on NSE's live-indices page are staged here on the identical criteria. The gaps, named rather than rounded away: three (Smallcap 500, REITs & Realty, Cement) are listed by the data source but it publishes no price series for them; Nifty500 Multicap 50:25:25 is not carried at all; India VIX is excluded deliberately — it is a volatility gauge derived from option prices, not a trend. Constituent lists are public for 76 of 82. The six without one are all ESG and Shariah indices (Nifty100 ESG, Nifty100 Enhanced ESG, Nifty100 ESG Sector Leaders, Nifty50/Nifty500 Shariah, Nifty Shariah 25) — their index pages publish a factsheet and a methodology but no constituent file, the screens being licensed. Those six are staged but membership is unknown, so no watchlist name can be placed in them.
• Volume is reported, not enforced — and the reason is worth stating. The correct test is volume in the breakout week, but the answer swings with a two-week change in where that week is dated: on the same four names the breakout-week reading is 0.8–1.5×, the best week within ±2 is 0.9–14.5×, and within ±4 it is 1.4–14.5×. Gating on any one of those would encode an arbitrary window as if it were his rule, so each name shows its breakout week and that week's volume, and you judge it.
• Blinding was imperfect — the Fisher and Lynch results and the holdings were known, and two names were charted before the criteria were frozen.
• Stage boundaries are a slope threshold, not a fact. ±2% over ten weeks separates “rising” from “flat”; names near that line can flip between runs, which is why transitions are reported rather than treated as certainties.