The BullBear Labs Guide

Why we exist

The shortest way to understand the product is to know which problem it was built for.

Most market apps do the same thing: they show a list of indicators, print “BUY” or “SELL” underneath, and stop there. That leaves three questions unanswered — Why? When would I be wrong? Is this even executable in my account?

Those three unanswered questions are where most losses come from. A “BUY” that does not say where it is wrong is not advice; it is a wish. And if a setup shows a 1:8 risk/reward ratio because the stop was placed inside the noise, that ratio is not information — it is an illusion.

What we built instead

BullBear Labs is a decision instrument, not a signal service. The instrument has three rules, and the entire product is built around them:

1. The engine produces the numbers; the AI interprets themEntry, stop, target, risk/reward and stop distance are computed deterministically server-side and sealed. The language model cannot change these numbers; it can only explain them. That is why the same analysis shows identical levels on your phone, on the web and in Telegram.
2. Every call carries its invalidationIf a plan is given, the level and the condition that break it are written too. A view with no invalidation condition is a view that cannot be tested.
3. We do not hide the measurementWhere a setup is weak, we say it is weak. If the stop sits below the noise floor, we flag it. If there is no setup, we say “no setup” — we do not invent numbers.

This guide explains what every reading on that instrument means. By the end you will know which line on an analysis card makes the decision, which line is context, and when you should simply close the card and move on.

The ecosystem: four surfaces, one account

Wherever you start, you see the same data and the same analysis. The surfaces are not copies of each other — each fits a different moment.

There is one account. A portfolio saved on your phone opens in the web terminal; an alert set on the web arrives on your phone. Premium bought on any surface applies to all of them.

The web is no longer a showcase

For a long time the website was a light promotional version: a few sample scans and a plain table. Not any more. The entire scan catalogue, the symbol detail board, favourites and alarms, the portfolio intelligence layer, the evidence card — whatever you do on mobile or in Telegram, you do in the web terminal too. The difference is not capability but SPACE: a wide screen shows the same information more densely, which is why the desktop is the most efficient surface when you are at a desk.

Screening: finding what to look at

Analysis is an expensive operation. Screening decides where you spend it. And screening is a far bigger part of this product than you might assume.

You cannot walk through thousands of stocks and coins by hand. A screen narrows the market by a measurable condition. But single-condition filters like “RSI below 30” are only the beginning — the real power sits in composite scans, where several conditions must hold at the same time.

Five categories

Scans built on investing styles

The fundamental category contains scans that apply the criteria of well-known investment approaches directly: Warren Buffett's criteria, Peter Lynch's valuation, Ben Graham's defensive stock screen. These do not say “look at this indicator”; they apply the numerical conditions of an investing philosophy across the whole market. For a long-term investor this is the most valuable part of the catalogue.

Composite scans

At the most selective end of the catalogue are composite scans that require several conditions to hold simultaneously. These can return nothing for days — that is design, not a defect. A selective scan that returns results every day is not selective enough.

• Bottom scans demanding perfect alignment: oversold, volume confirmation and a reversal structure all at once
• Volatility singularity: sudden expansion after a long compression — aiming to catch the moment of the break
• Bear-trap scans: structures that looked like breakdowns but recovered, carrying the fingerprint of institutional buying
• Flawless momentum: candidates where trend, acceleration and volume line up together

Using screening well

1. Pick a category that matches your approach. If you follow trends, an oversold list will keep walking you into falling knives.
2. Start with a single-indicator filter; if the result is too crowded, move to a composite scan.
3. Sort by the columns; the top row is not automatically the best candidate.
4. Add an interesting symbol to your watchlist first, rather than analysing it immediately. Watching it for a day eliminates most bad entries.
5. Assign the scan you like to the Opportunity Agent as its strategy — so you do not have to check by hand every day.
6. Run the analysis only once the question is clear. Your analysis quota is finite; spend it on the right question.

Scans run on Borsa İstanbul, the US exchanges, the UK, Germany and the major crypto venues. In crypto the fundamental category is naturally disabled.

Analysis: asking the right question

The quality of an analysis depends on the context you give it. Choosing the timeframe and the mode is the question itself.

What a timeframe actually means

The timeframe answers “at which scale do you want this to be true?” A setup that looks strong on a 15-minute chart may be a small bounce in the middle of a falling daily trend. Both readings are correct; they answer different questions.

God's Eye: five timeframes, one call

Most setups that look strong on a single timeframe fall apart once you look at the one above it. God's Eye scans five timeframes together and measures how much they agree. When all five point the same way, that is far stronger than the same signal seen on one timeframe alone.

Low agreement is not a defect — it is information. “Short term up, medium term down” tells you the setup is short-lived and that you should shrink the target accordingly.

Spot and futures mode

For crypto symbols you are asked for a trading mode. This is not cosmetic: with a spot account you cannot profit from a falling market. When the mode is “spot” and the direction is down, the product will not produce a short plan you cannot execute — it gives you a plan to protect what you hold instead. Details in the next section.

What the analysis sees

For transparency: the engine hands the language model the price and volume data of the last CLOSED bar on the selected timeframe, computed technical indicators (RSI, MACD, ADX, ATR, moving averages, Bollinger, MFI, VWAP, pivots), fundamental ratios where available, recent headlines, exchange filings, and market-regime measurements. The unclosed bar is deliberately excluded: an indicator computed on half a bar produces a wrong result.

The evidence scene: what you watch while waiting

While the analysis runs, the screen shows the symbol's own chart rather than a spinner. The detected evidence is drawn on it in sequence: pattern geometry (shoulders, neckline, measured target), indicator divergences, the Fibonacci leg and its levels, volume zones. All of it is computed before the AI writes a single word — because it is exactly the evidence the model reads. The double bottom you watch and the double bottom you read about are the same object.

When the result arrives the scene does not close; it becomes the Evidence card. There you can switch layers off and on, add indicators, and replay the reel. The goal is not to convince you but to let you check.

How Fibonacci levels are drawn

Fibonacci levels are measured from the last swing leg, not from a fixed window: the move between two pivots (shown as a dashed line) defines the 0% and 100% ends, and the ratios in between give the retracement levels. If the leg is up, the retracement is measured down from the high; if it is down, up from the low — which is why 38.2% sometimes sits above 61.8% and sometimes below. That is correct: the measuring direction follows the leg.

Volume is not a single number

A bar's volume is split between buyers and sellers according to where the close sits inside the candle range. That makes the real question answerable: is the advance backed by buying, or is it drifting up on seller volume? When cumulative delta diverges from price, that is a sign of distribution or accumulation — if price makes a new high but buying power does not, the advance is hollowing out.

The same calculation yields money-memory zones: the bars with the highest buy and highest sell volume in the lookback window. That bar's price range is drawn as a box; price returning there is expected to react.

Channels, wedges, triangles

Some patterns are not a single event but a GEOMETRY that price is squeezed inside. Two trendlines are drawn — one through the highs, one through the lows — and the shape’s name follows from the slope of those two lines and how the gap between them changes. Both lines in the same direction with a constant gap is a channel; same direction with a narrowing gap is a wedge; one flat and one sloped is a triangle.

Directional claims belong only to CONTRACTING shapes: rising wedge bearish, falling wedge bullish, ascending triangle bullish, descending triangle bearish. We do not paste the same label on their expanding siblings — there the information is volatility, not direction, and the stop should be widened accordingly. A symmetrical triangle carries no direction either: it is a coiled spring, and the energy releases whichever way it breaks.

A break is confirmed by a CLOSE, not a wick; a candle that pierces the line and comes back is not a break. If the break has gone stale (more than a few bars old) the shape is not mentioned at all — it is history, not evidence.

Structure breaks and order blocks

What we call “structure” is the chain of swing highs and lows. When that chain flips — when a new low runs clearly below the prior low, or a new high clearly above the prior high — structure has broken. “Clearly” is a measured thing here: the move must travel a third of the broken leg beyond the level. Without that buffer, structure flips on every bit of noise and means nothing.

The last opposite candle before the breaking move is drawn as an “order block”: the last down candle before an upward break, the last up candle before a downward break. The assumption is that this is where the large order was left, so if price returns there the same side is tested again. Zones left behind by the previous leg that have switched sides are marked separately as “breakers” — former resistance turned support, or the reverse.

Reading the result

The heart of this guide. What every line on the analysis card says, and what you should decide from it.

4.1 · The action badge — “what do I do?”

The badge at the top of the card is the primary answer. It says “ENTER”, “WAIT FOR PULLBACK”, “HOLD”, “REDUCE”, “EXIT” or “STAND ASIDE”. Note that this is a different axis from the market view. The view on a symbol can be “up” while the action is “wait for a pullback” because price is already extended. Direction and timing are not the same thing.

4.2 · The action matrix — three readers, one analysis

Three different people look at the same symbol, and you cannot tell them the same thing:

• Someone holding it: “do I sell, hold, or add?”
• Someone waiting in cash: “is this a place to enter?”
• Someone who can go short: “is this a short setup?”

The card answers all three separately. In systems with a single badge, “STAND ASIDE” was read by holders as “sell, get out” — when what it meant was “not suitable for a new entry”. Making that distinction is one of the most useful things the card does.

4.3 · Plan type — a plan you can actually execute

4.4 · Levels: entry, stop, target

Three numbers form the skeleton of a plan. Entry comes in two forms: a market entry (at the current price) or an entry anchored to a zone. In the second case the plan is conditional — nothing starts until price reaches the stated zone, and the stop and target are computed relative to that zone. The card says so explicitly; executing a “pending” plan immediately means executing a different plan.

4.5 · R:R and the noise floor — the most important line

Risk/reward is the distance to target divided by the distance to stop. 1:3 means that when you are right, you expect to make three times what you lose when you are wrong. It is the number everybody looks at, and the number that most easily fools them.

Here is why: move the stop closer to price and the ratio grows automatically. A 1:8 computed from a 0.3% stop looks wonderful — but if that stock already swings 2% on an ordinary day, the stop sits inside normal intraday movement and will be swept even when the direction was right.

A practical reading: 2×ATR and above is healthy. Between 1 and 2×ATR calls for care. Below 1×ATR deserves suspicion no matter how attractive the R:R looks.

4.6 · Setup grade — the letter and its reasoning

A setup is graded A, B, C or F. The letter alone is not information, so the reasoning is written next to it: “strong R:R”, “weak volume support”, “stop is not structural”. A low grade does not suppress the plan — but sizing down on a low-graded setup is the correct behaviour.

4.7 · The confidence figure

Confidence shows how strongly the model backs its own call. It is not a probability: 78% confidence does not mean “you win 78% of the time”. It summarises how well the evidence agrees with itself. Low confidence tells you the setup carries contradictory signals.

4.8 · Invalidation

The level and the condition that break the plan are written down. This is the least-read and most valuable line on the card: knowing in advance when to stop defending an idea is the only reliable way to keep losses small. Trying to define the moment you were wrong while you are already in the position is almost always too late.

4.9 · Horizon

Every call holds for a period, and the card states it in trading days: “this call is for 5–8 trading days”. The horizon does not come from a fixed table; it is derived from the symbol's volatility and the geometry of the stop. A wide-stopped, slow setup has a long horizon; a tight, fast one has a short horizon.

4.10 · The Devil's Advocate

The card contains two arguments the model produced AGAINST its own thesis. That is deliberate design: defending your own view is easy, constructing the counter-argument is hard. If reading the counter-arguments changes your mind, you should not have taken that position.

4.11 · Entry quality and extension

A symbol's direction can be right while price is already stretched. Extension is measured on a separate axis and stated separately. “Direction is up but it is late to enter” is not a contradiction; it is the most common situation, and it is why the card says “wait for a pullback”.

4.12 · Reading the God's Eye table

Five timeframes are listed with their directions. Read it this way: higher timeframes give context, lower ones give timing. If the higher frames are down and the lower ones are up, you are holding a bounce — keep the target small. If all of them align, the setup is strong and permits a wider target.

The Evidence Board: what drives the verdict

An analysis is not an opinion but a chain of reasoning. This section opens that chain: which professional system looked at what, what it found, and how it touched the decision.

Every analysis scans 13 professional systems. Some find something, some do not — and finding nothing is a measurement too. Each finding is read the way that tool is read in the literature, and the result surfaces in three places: the evidence strip under the verdict, the list of reasons behind the decision, and the evidence card drawn over the chart.

The location condition — the most-skipped detail

Most of these tools are conditional. An order block carries information when price RETURNS to it; a zone five ATR away is not “support”. The “buy the floor, sell the ceiling” rule belongs only to parallel channels, never to a contracting triangle. A rule that does not ask about location makes evidence look stronger than it is — so every item asks where price stands relative to that structure.

Confirmation, contradiction, and the effect on the verdict

Each item either confirms the signal, contradicts it, or carries context. This is not a scoring contest: the board does not PRODUCE direction, it tests the existing verdict. Confirmation raises confidence; contradiction both lowers confidence and — if strong enough — steps the signal down one rung (“BUY” → “CAUTIOUS BUY” → “WAIT”). It never steps a signal up and never flips it to the other side.

Break, retest, and the failed breakout

A line breaking is the start of the story, not the end. The classic sequence: price breaks, moves away, returns to test the boundary, and moves away again — “test and go”, the strongest confirmation in this family. If price comes BACK INSIDE the channel after breaking, that is a failed breakout; the side that made the move could not sustain it, and classically that is a signal in the OPPOSITE direction. If price crossed the whole channel in one move, that is not a breakout but a volatility expansion — widen the stop accordingly.

Trend speed: a comparable angle

Saying “this trend is steep” means nothing without a unit. Price is divided by volatility into a dimensionless space, and the trend’s angle is measured there: 45 degrees means one standard deviation per bar. That angle is independent of instrument and timeframe — the same number means the same speed on a BIST stock and on a crypto pair.

Multi-scale consensus — is the trend robust?

The same chart is read at six resolutions: short blocks see the latest wave, long blocks see the main structure. If they all point the same way, the trend rests on STRUCTURE rather than a single wave. If they disagree — short blocks up, long blocks down — the trend exists only at the scale you happen to be looking at, which is fragile ground. The information here is not direction but ROBUSTNESS, so disagreement counts as a warning whichever way it points.

The evidence card: verify the claim with your own eyes

Every structure you read in the text is drawn on the symbol’s own candles in the evidence card: pattern geometry, indicator divergences, the Fibonacci leg, order block boxes, money-memory zones. You can toggle layers one by one, pick a different oscillator in the lower panel, and replay the reel. The point is not to convince you but to let you CHECK — because once you can verify a claim, you decide for yourself whether to trust it.

Position awareness

Enter your portfolio and the analysis becomes personal. The same symbol speaks differently depending on your cost.

The portfolio intelligence layer

The portfolio screen is no longer a list but a reading. Four measurements sit above it: today’s profit and loss, the strongest and weakest position, a concentration warning, and the market breakdown. None of them is a forecast — every number derives from the value of your own positions.

The concentration row warns when a single position passes 40% of the portfolio. That is a measurement, not a forecast: if you are wrong on that asset, the whole portfolio is wrong. When everything is in profit the label reads “weakest” rather than “biggest loser” — the label follows the fact, not the other way round.

Once you add a holding to your portfolio, the product knows your cost and your size. A personal block appears on the analysis card: your cost, your current value, your unrealised profit or loss, and an action recommendation based on them.

What it takes into account

• Whether your cost sits at support or at resistance — a cost stranded above resistance means a trapped position
• How far breakeven is in ATR terms — saying “it will recover” requires that distance to be realistic
• The state of your unrealised P/L: in profit, at breakeven, underwater, or deeply underwater

Personal tools and agents

An analysis is a moment; investing is a process. The real difference of this product is the part that works while you are not looking — three agents, alerts, portfolio and the briefing.

Most market apps work when you open them. This is where BullBear Labs separates: the agents you configure watch the market while you sleep and speak only when there is something to say.

🎯 The Opportunity Agent

It scans the market in the background with the strategy you choose, and notifies you when the conditions form. The strategy is picked from the scan catalogue: golden cross, oversold reversal, momentum break, post-squeeze expansion — whichever matches your approach.

A critical detail: candidates are found by the mathematical decision engine, not by the language model. That drives the cost to zero and makes the result reproducible. If there are no candidates, nothing is sent — we do not send a notification to say “no opportunities today”, because an empty notification devalues the next real one.

📡 The News Agent

It watches the global news flow, analyses breaking developments with AI and identifies bullish or bearish sentiment. Instead of a pile of raw headlines, it tells you what the news means for the symbols you follow. Official exchange filings are tracked separately; a company's own disclosure and a news site's commentary are not weighted the same.

💼 The Portfolio Agent

At the day and hour you choose, it analyses your entire portfolio and produces a report. You can pick at most three days a week — a deliberate limit: a report that arrives every day stops being read. The report assesses the portfolio as a whole rather than symbol by symbol: concentration, correlation risk, and which position needs attention.

One-tap alerts from an analysis

When the card proposes a level, you can turn it into an alert with one tap. You do not compute the level — the correct one for the plan type (zone, stop or exit) has already been determined. That removes the problem of ending up with four different alerts on four different screens. When the alert fires, the notification also says which plan the level came from.

Watchlist and portfolio

The watchlist holds the symbols you follow; the portfolio holds what you actually own, with cost and size. They serve different purposes: the watchlist is the agents' hunting ground, while the portfolio is what makes an analysis personal. Once you enter a portfolio, the analysis card starts speaking relative to your cost (see Section 05).

The daily briefing

At the hour you choose, a summary built from your watchlist and portfolio arrives: overnight moves, news that concerns you, and positions that need attention. With an empty portfolio and an empty watchlist, no briefing is sent.

Quiet hours

You can define a quiet window so notifications do not arrive while you sleep. The window can follow your device's local time, so it does not drift when daylight saving changes.

Comparison and the thesis ledger

You can place several symbols side by side and compare them on the same criteria — that is the correct way to ask “which one should I buy?”. Your analyses are also saved; you can look back at what was thought about a symbol earlier and what that call turned out to be. Seeing your own past decisions teaches more than any single analysis.

Risk and realistic expectations

This section does not sell the product. But we would rather you did not take a position without reading it.

A hit rate alone says nothing

“60% of the signals worked” is meaningless on its own. What matters is the net edge: hit rate minus the market's base rate. In a rising market, 60% of randomly chosen stocks also rise. A tool is worth what it adds ON TOP of the base rate.

Position size matters more than entry

The card gives a position-size suggestion, written as a percentage of your capital. The reason is simple: a right entry with the wrong size loses more than a wrong entry with the right size. Sizing down on low-graded setups is easier and more effective than trying to raise your hit rate.

What this product cannot do

• It cannot know where price will go. Nobody can.
• It cannot foresee news and event risk; an earnings surprise, a regulatory change or a macro shock is outside the model's field of view.
• It does not know your risk tolerance. The recommendation is about the state of the setup, not about your situation.
• On very illiquid symbols, the levels it produces may not be executable.

Ten common mistakes

The patterns we see repeated most often, and which cost the most.

1. Treating a screener hit as a signal. Screening produces candidates, not decisions.
2. Taking a position without reading the invalidation line. Entering without knowing where you are wrong leaves the exit to emotion.
3. Celebrating a high R:R without checking the stop distance. A stop below 1×ATR makes that ratio a lie.
4. Entering at market when the card says the entry is anchored to a zone. The plan is conditional; executed before the condition, it is a different plan.
5. Choosing the timeframe by which answer you like rather than by the trade you intend.
6. Turning on God's Eye for every symbol. Save the resource for the one you will actually decide on.
7. Treating an old analysis as still valid after its horizon has expired.
8. Sizing a C-graded setup the same as an A-graded one.
9. Skipping the Devil's Advocate section. It is the only part that can change your mind.
10. Not setting quiet hours and then making a sleepy decision on a notification at 3am.

Glossary

Short, honest definitions of the terms that appear on the cards.

RSI

Measures the speed of recent moves on a 0–100 scale. Below 30 is read as “sold hard”, above 70 as “bought hard”. In a strong trend it can stay pinned at an extreme for a long time; on its own it is not a reversal signal.

MACD

The difference between two moving averages. Crossing above zero shows momentum turning up, below zero turning down. It produces frequent false signals in a sideways market.

ADX

Measures the STRENGTH of a trend, not its direction. Above 25 means a strong trend, below 20 a sideways market. Trend strategies do not work in a sideways market.

ATR

Average true range: how much the symbol typically moves in one bar. It is the correct unit for judging whether a stop distance is meaningful.

EMA

Exponential moving average; weights recent prices more heavily. The 20, 50 and 200 are the most widely watched and are treated as support and resistance.

Bollinger bands

An envelope showing how far price has strayed from its average. Narrowing bands signal compression; widening ones signal rising volatility.

MFI

A volume-weighted RSI. It measures the direction of money flow, and is unreliable when volume data is missing.

VWAP

Volume-weighted average price. Used as a “fair price” reference in intraday trading.

R:R

Risk/reward. Distance to target divided by distance to stop. If the stop distance is below the noise floor, the ratio loses its meaning.

Consensus

The shared tendency of several indicators. More robust than any single indicator, but lagging.

Regime

The market's current character: trending, compressed, or volatile. The same indicator means different things in different regimes.

Confluence

Different methods pointing at the same level. If a moving average, a pivot and a volume profile all mark one level, that level is stronger.

Frequently asked questions