TDU Auto Fibs: Fibonacci levels without drawing every swing

Auto Fibs draws every meaningful retracement on the chart for you, ranks them, stacks them into confluence clusters where several swings agree on the same price, and quietly removes the ones price has already broken. You stop maintaining drawings and start reading a live map of where pullbacks actually end.

Every discretionary trader who has ever used a Fibonacci tool knows the same two frustrations. The first is labour: you have to find the swing, anchor the tool, decide whether the wick or the body is the real high, and then redo the whole thing twenty minutes later when a new leg prints. The second is bias: because drawing takes effort, you draw the swing that supports the trade you already want to take. Both problems disappear when the levels are generated by a rule instead of by your mouse.

TDU Auto Fibs is the retracement engine in the SabrTrader premium suite. It scans price for qualifying swings using a strength filter and a minimum swing size, builds a retracement set on each one, keeps up to thirty of them in memory, and then decides which ones deserve screen space right now. Where several independent retracements land inside the same narrow band of price, it marks a cluster \u2014 the single most useful piece of information a fib tool can give you, because a level three swings agree on is not the same animal as a level one swing agrees on.

What is TDU Auto Fibs, exactly?

It is a chart study that automates the entire lifecycle of a Fibonacci retracement. Instead of one manually anchored tool, you get a rolling population of retracements built from objective swing pivots. Each set is anchored from the origin of an impulse (100%) to its termination (0%), and each set draws the retracement percentages you configure \u2014 by default 50, 61.8 and 65 \u2014 plus optional extension targets projected beyond the impulse at -23.6 and -61.8.

Because the swings are found by rule, the levels are reproducible. Two traders running the same settings on the same instrument see the same lines. That matters more than it sounds: a level only has power when enough participants can see it, and a rule-based level is far closer to what an algorithm or another desk would compute than a hand-drawn one anchored to a wick you happened to like.

The second half of the design is subtraction. A naive auto-fib script produces spaghetti \u2014 forty sets of lines, every one of them technically valid, none of them readable. Auto Fibs solves that with three mechanisms working together: relevance ranking (show only the N most relevant retracements), auto-hide (retire sets that no longer describe current price), and broken-fib removal (delete a level once price has traded a defined number of ticks through it). The result is a chart that usually carries two or three live retracement sets and one or two clusters, not a plate of noodles.

Why manual Fibonacci drawing quietly costs you money

Consider a normal intraday session. Price makes an impulse leg, retraces, makes a higher low, extends, then chops sideways, then extends again. That is four or five legs before lunch. To keep an honest fib map you would need to anchor a tool on each of them, keep the older ones for context, and remember which ones are still valid. Nobody does that consistently for six hours. What actually happens is that you draw one fib on the leg you are trading, take the 61.8 entry, and never notice that the same 61.8 sat two ticks away from the 50 of the larger leg and the 65 of the leg before that \u2014 a triple confluence that would have justified twice the size.

The inverse error is worse. You draw a fib, price blows through 61.8 by fifteen ticks, and you keep the drawing on the chart because deleting it feels like admitting you were wrong. Ten minutes later price returns to that line and you treat a broken level as support. Auto Fibs enforces the discipline you would enforce on yourself if you had infinite patience: once a level is broken by TickBreak ticks, it stops being drawn, so you cannot lean on it.

There is also the question of what you do not see. A manual tool shows one leg. The market is composed of nested legs on several scales at once, and the highest-probability turns tend to occur where scales agree. You can only see agreement if all the scales are drawn simultaneously. That is the structural argument for automation, and it is why the cluster logic \u2014 not the line drawing \u2014 is the real product here.

How does the engine find swings?

Everything starts with pivot detection. Two parameters govern it.

Strength (default 10) defines how many bars must sit on each side of a candidate pivot without violating it before that pivot is accepted as a swing point. A strength of 10 means a swing high needs ten bars to its left and ten to its right that failed to exceed it. Lower values produce more, smaller swings and therefore more retracement sets; higher values produce fewer, larger, more structural swings. On a fast 1-minute or volume-based chart you might drop to 5\u20137 to capture the micro-legs a scalper trades; on hourly and daily charts you might push to 15\u201320 so that only genuine structure qualifies.

MinSwingTicks (default 20) is the size filter. A swing must span at least this many ticks before the engine bothers to build a retracement from it. This single number does most of the anti-noise work, because a two-tick wiggle that technically satisfies the strength test carries no information about where a pullback will end. Set it relative to the instrument's typical range: a value that filters correctly on an index future will be far too small on a high-tick-value crypto pair and far too large on a low-priced stock. The practical rule is to set MinSwingTicks to roughly one third of the average leg you actually trade.

MaxRetracements (default 30) caps how many retracement sets the engine tracks. Thirty is generous \u2014 it gives the cluster logic plenty of raw material to work with while keeping computation bounded on long histories. You rarely need to change it, but on very long intraday sessions with heavy legging you may raise it so that morning structure is still available in the afternoon, and on a slow daily chart you can lower it because thirty daily swings reach back further than is useful.

Confluence clusters: the feature that changes how you size

Set ShowCluster to true and Auto Fibs looks across every live retracement set for prices that agree. When two or more levels from different swings land within the tolerance defined by ClusterWidth (default 100), the engine marks that band as a cluster instead of leaving you to eyeball whether three dotted lines are \"close enough\".

Why does this matter? Because confluence is the closest thing retracement analysis has to a probability filter. A lone 61.8 is a coin flip with a nice story attached. A zone where the 61.8 of the current leg, the 50 of the parent leg and the 65 of last session's leg all fall inside a few ticks is a place where three different groups of participants have a reason to act at the same price. Those are the pullbacks that hold on the first touch, and those are the pullbacks where you can justify a wider stop and a larger position because the invalidation is genuinely structural.

ClusterWidth is the tolerance dial. Raise it and more levels merge into fewer, wider clusters \u2014 useful on volatile instruments where a level is really a zone. Lower it and you demand tight agreement, producing fewer clusters but sharper ones. The default of 100 is deliberately generous so that clusters are visible out of the box on a wide-ranging instrument like BTC; on tighter, smaller-tick markets you will normally reduce it substantially until the clusters it paints look like the zones you would have drawn by hand.

Auto-hide, relevance and broken fibs

Three parameters keep the chart honest.

AutoHide (default on) tells the engine to display only the retracement sets that currently matter, rather than all thirty it is tracking. RelevantFibCount (default 3) sets how many. Three is a well-chosen default: it typically gives you the retracement of the leg you are trading, the retracement of its parent leg, and one older set that still contains price. That is enough context to see nesting without turning the chart into a grid.

Raise RelevantFibCount to 5 or 6 when you are hunting clusters and want maximum raw material \u2014 more sets means more chances for agreement, at the cost of visual density. Drop it to 1 or 2 when you are scalping and want a single clean pair of lines to trade against. If you turn AutoHide off entirely, every tracked retracement draws; that is occasionally useful for a research pass in replay, and almost never useful live.

HideBrokenFibs (default on) and TickBreak (default 2) handle invalidation. A level is considered broken once price trades TickBreak ticks beyond it, at which point the line is removed. Two ticks is a tight, honest threshold: it tolerates a one-tick overshoot on the print but does not let you pretend a level held after a clear penetration. Widen TickBreak if your instrument has noisy spreads or you deliberately want tolerance for stop-run wicks; keep it tight if you want the chart to reflect strict level violation.

This is where automation beats discipline. Nobody deletes their own broken levels in real time. Auto Fibs does it every bar, which means the levels remaining on your chart at any moment are the ones that have never been violated \u2014 a fundamentally different information set from \"levels I drew at some point today\".

Why 50, 61.8 and 65 \u2014 and what the negative targets mean

The default retracement set is FibRetracementLevel1 = 50, FibRetracementLevel2 = 61.8 and FibRetracementLevel3 = 65, with levels 4 through 8 set to 0 (inactive). That is not laziness; it is a deliberate, tested trade zone.

The 50% mark is not a Fibonacci ratio at all \u2014 it is the midpoint, and it works because it is where the average participant in the impulse leg is at breakeven. The 61.8% level is the golden ratio retracement, the most watched number in the family and the one where continuation trades cluster. The 65% level is the sleeper: it sits just beyond 61.8, which is exactly where stops from 61.8 entries live. Having 65 drawn means you can see the difference between \"deep pullback that held the golden ratio\" and \"pullback that swept the golden-ratio stops and then held\". Those are two distinct setups with different entries, and most fib tools blur them into one.

Levels 4\u20138 are yours. Set FibRetracementLevel4 = 38.2 if you trade shallow pullbacks in strong trends, add 78.6 for deep flush reversals, add 23.6 for momentum continuation on very strong legs. Anything you set to 0 simply does not draw, so you can run a three-level minimalist chart or an eight-level research chart from the same indicator.

The targets work on the same percentage scale, extended past zero. FibTarget1 = -23.6 and FibTarget2 = -61.8 project beyond the end of the impulse in the direction the impulse was travelling. In plain terms: if 100% is the origin of the leg and 0% is where the leg ended, then -23.6% is a modest extension past the prior high and -61.8% is an aggressive one. These are your take-profit shelves for continuation trades \u2014 you enter in the 50\u201365 retracement zone and you know before you click where the first and second scale-outs sit. FibTarget3 through FibTarget8 are free slots; -100 and -161.8 are the natural additions for traders who hold for full extensions.

How do I read it on the chart?

Work top-down, in this order, every time. The sequence matters because it stops you from finding a level first and a reason second.

1. Identify the dominant leg. The widest live retracement set on the chart is your context. If price is in the upper third of that set, you are in a trend pullback; if it is below the 65 line, the leg is under threat and you should treat continuation trades as counter-structural.
2. Find the cluster. If a cluster band is visible, that is your primary decision zone. Everything else is secondary. A cluster is the engine telling you multiple scales agree.
3. Check what is still alive. Remember that broken levels have been removed. If the chart looks unusually empty above price, that is information: the levels there have already failed.
4. Read the labels. With ShowPrice and ShowPercentage both on, each line tells you the exact price to work an order and the percentage so you know whether you are buying a shallow or a deep pullback. The percentage is the context; the price is the order.
5. Locate the targets. Before you enter, note where -23.6 and -61.8 sit. If the distance from your entry to -23.6 is smaller than the distance from your entry to the cluster's lower edge, the trade is not worth taking. Auto Fibs makes that comparison visual instead of arithmetic.

One habit worth building: treat the 0% and 100% anchors as levels in their own right. They have their own colour, width and style settings (StrokePercent0* and StrokePercent100*) precisely because they are structural \u2014 100% is the point that must not be lost for the leg to remain intact, and 0% is the breakout line that turns a pullback trade into a continuation trade. Style them solid or brighter than the interior levels so your eye separates structure from retracement.

The structural support and resistance module

Auto Fibs ships with an optional second engine: horizontal support and resistance zones derived from price structure rather than from Fibonacci maths. It is off by default (ShowSupportResistance = false) so that new users see a clean fib chart first, but it is worth turning on once you are comfortable, because fib confluence plus a real historical level is a stronger argument than either alone.

SupportResistanceStrength (default 30) is the pivot strength used to qualify a structural level \u2014 much higher than the fib swing Strength, because you want only major turning points to generate zones. SupportResistanceTicks (default 8) sets the thickness of each zone: a level is a band, not a line, and eight ticks is a sensible starting width that you should scale to your instrument. SupportResistanceTouches (default 0) filters by how many times price has interacted with the level; at 0 the touch filter is not applied, so you see every qualifying zone. Raise it to 2 or 3 when you only want levels with a proven history of reaction. AutoHideSR (default on) retires zones that are no longer relevant to current price, using the same philosophy as the fib auto-hide.

Styling is independent: SupportResistanceColor defaults to a translucent tan (semi-transparent by design, so zones sit behind candles), with SupportResistanceWidth = 3 and a dotted style. Keep the zones dimmer than your fib lines. The visual hierarchy you want is: candles brightest, fib cluster next, individual fib lines next, S/R zones faintest \u2014 background context, not foreground signal.

Four playbooks you can trade tomorrow

Playbook 1 \u2014 Cluster Continuation

Context: a clear impulse leg has printed and price is pulling back into a visible cluster band. Trend structure on the higher timeframe is intact.

Entry: work a limit order at the far side of the cluster, not the near side. If the cluster spans a band, the near edge is where impatient buyers sit and the far edge is where the fill quality is. If you need confirmation, wait for a bar to close back inside the cluster after tagging its far edge.

Stop: below the cluster's outer boundary plus your instrument's normal noise allowance \u2014 in practice, below the deepest fib line contributing to the cluster. If the cluster's lowest contributor is the 65 of the dominant leg, your stop belongs beneath that, not beneath 61.8.

Invalidation: the moment the deepest contributing level is broken by more than TickBreak ticks, Auto Fibs deletes it. That deletion is your signal that the premise is gone \u2014 if you are still in the trade with no line beneath you, you are trading hope.

Targets: scale at -23.6, trail the remainder toward -61.8. Move the stop to the 0% anchor once -23.6 prints.

Playbook 2 \u2014 The 65 Sweep

Context: price pushes through 61.8 and trades into the 65 line. This is the setup the third default level exists for.

The logic is behavioural. Thousands of traders enter at 61.8 with stops just beyond it. Price sweeping into 65 collects those stops, which produces a burst of selling into an area where larger buyers are waiting. If the leg is genuinely healthy, price reverses from 65 hard and fast, often reclaiming 61.8 within a bar or two.

Entry: on the reclaim of 61.8 from below, after the 65 tag. Not on the tag itself \u2014 the tag is where you get run over.

Stop: beneath the 65 line by a fixed tick buffer. If 65 does not hold, the retracement has become a reversal and the whole leg is in question.

Invalidation: a close below 65, or the disappearance of the 65 line from the chart. Note that when HideBrokenFibs is on, the line vanishing is the invalidation \u2014 you do not need a separate rule.

Targets: the 0% anchor first (a full retracement of the pullback), then -23.6.

Playbook 3 \u2014 Failed Level Flip

Context: a level breaks, Auto Fibs removes it, price retests the void from the other side. This is the short playbook of the set, and it is the most underused.

When the 61.8 of a dominant leg fails and price returns to it from below, the previous demand area becomes supply. Because broken fibs are hidden, you will not see a line there \u2014 which is why you should note the price of the level that just disappeared. The empty space above price where lines used to be is a map of failed support.

Entry: short on the retest of the broken level's price, with rejection confirmation. Best when the broken level also sat inside a cluster \u2014 a cluster that fails is a strong statement, because multiple scales just got proven wrong at once.

Stop: above the next surviving fib line or the 0%/100% anchor above, whichever is closer.

Targets: the next live cluster below, or the 100% anchor of the failed leg.

Playbook 4 \u2014 Nested Agreement Reversal

Context: price is deep in the retracement of a large leg (past 65) while simultaneously completing a full extension of a smaller counter-leg into a target level. Two scales, opposite directions, same price.

This is the highest-conviction, lowest-frequency setup Auto Fibs produces, and you can only see it because multiple retracement sets are drawn at once. Raise RelevantFibCount to 5 when you are specifically hunting it. Add a structural S/R zone in the same band and you have three independent reasons at one price.

Entry: scale in across the band rather than at a single price, because deep reversals rarely give you a clean single touch.

Stop: beyond the 100% anchor of the dominant leg. If that goes, the trend is over and you should be trading the other direction, not defending a fib.

Targets: back to the 0% anchor of the dominant leg, scaling at each surviving fib line on the way up. Because Auto Fibs deletes levels as they break, every line that vanishes above you as price advances is a confirmation that the move has real participation.

How do I combine it with the rest of the suite?

Fibonacci levels answer \"where\". They do not answer \"who\". The strongest workflow in SabrTrader layers Auto Fibs with a volume-based tool that shows participation at the level.

With Volume Profile: the highest-quality clusters are the ones that land on a high-volume node or on the value-area edge of the previous session. Price coming back into a cluster that sits on a naked point of control is a pullback into an area that has already proven it can absorb size. Conversely, a cluster in a low-volume gap is fragile \u2014 nothing there has to defend it.

With Footprint: the cluster gives you the price; footprint gives you the tape at that price. You want to see selling absorbed \u2014 heavy sell prints with no downward progress \u2014 inside the cluster band before you commit. Delta divergence at the far edge of a cluster is one of the cleanest entry triggers available, because you have a rule-based location and an order-flow confirmation at the same time.

With Market Structure: use it as the permission layer. Trade cluster continuation only in the direction of the prevailing structure, and reserve deep-retracement reversal trades for when structure has already shifted. This one filter removes most of the losing trades a pure fib approach generates.

With TDU Price Action: pattern confirmation inside the zone. A cluster plus a reversal pattern printed inside it is a two-factor entry, and because both are rule-based you can review them honestly in replay afterwards.

Works on every market you can chart

Auto Fibs is instrument-agnostic. It reads price and ticks, so anything with an order book and a tick size works. What changes between markets is calibration, not method.

Futures. Index, energy and metals futures are the natural home for retracement trading because sessions have structure: an opening drive, a pullback, a continuation. Set MinSwingTicks so that only legs worth trading qualify, and pay attention to clusters that coincide with the prior session's high or low. On instruments where a tick is small relative to the daily range, expect to raise MinSwingTicks well above the default 20.

Stocks. Retracements behave well on liquid large caps and on anything with a clean intraday trend. Because tick size is uniform but price is not, MinSwingTicks and ClusterWidth need per-symbol adjustment \u2014 a 20-tick swing means something entirely different on a $15 stock than on a $600 one. Save separate templates for your low-priced and high-priced watchlists.

Crypto. The BTC/USDC example on this page is not accidental \u2014 crypto is where retracement analysis earns its keep, because the market trends 24/7 with no session reset and legs run further than in equity index products. The generous default ClusterWidth of 100 exists precisely so clusters remain visible on wide-ranging crypto pairs. Watch for clusters forming around round numbers; when a fib cluster and a psychological level coincide in crypto, reactions are unusually violent.

FX and CFDs. Same logic, tighter ranges. Reduce ClusterWidth substantially and consider a lower TickBreak, since noise around a level is smaller in absolute terms.

Complete settings reference

Display and levels

Swing detection and relevance

Invalidation

Retracement levels and targets

Support and resistance zones

Styling

Recommended presets

How it compares

The honest comparison is not against other Fibonacci indicators \u2014 it is against the manual tool you are using now and against the free auto-fib scripts that circulate on every platform.

The difference in one sentence: a manual tool gives you one opinion, a free script gives you one opinion automatically, and Auto Fibs gives you a ranked, self-cleaning population of opinions with the agreements marked.

Common mistakes to avoid

Trading every level. The chart draws three levels per set and up to three sets. That is not nine trades. The tradeable locations are clusters and, occasionally, a clean single level in an obvious trend. Everything else is context.

Leaving MinSwingTicks at the default on every instrument. Twenty ticks is a starting point, not a universal truth. On some markets it will produce a retracement on every wiggle; on others it will produce almost nothing. Calibrate once per instrument and save a template.

Turning AutoHide off because you want to see everything. You do not want to see everything. Seeing everything is how you find a level to justify any trade in any direction. The relevance filter is a discipline mechanism disguised as a display option.

Ignoring the disappearance of lines. When HideBrokenFibs removes a level you were leaning on, that is a live signal, not a rendering quirk. Build the habit of noticing what left the chart.

Using targets as guarantees. The -23.6 and -61.8 projections are reference shelves for scaling, not price predictions. Their job is to make your reward-to-risk visible before entry.

Setting it up in SabrTrader

Install the platform, open a chart of the instrument you trade, add TDU Auto Fibs from the indicator list, and accept the defaults for the first session. Watch how the levels behave without trading them. On the second pass, adjust in this order: Strength to match your timeframe, MinSwingTicks to match your instrument, ClusterWidth until clusters look like zones you would have drawn, then RelevantFibCount to match your tolerance for information density. Leave the invalidation settings alone unless you have a specific reason.

Then use replay. Load a week of history, run it forward bar by bar, and log what happened at each cluster. You will discover very quickly which clusters on your instrument are worth a limit order and which are worth waiting for confirmation. That log \u2014 not the settings \u2014 is what makes the indicator profitable for you.

Frequently asked questions