OPEN-SOURCE SCRIPT
Market Dictionary [GBB]
Can price be translated into a language?
I wondered if it could. So I built a dictionary for the market. This
indicator is that dictionary — including everything it does NOT say.
How it reads the chart: Every bar, the recent closes are reduced to
their shape and spelled as a short word — a few letters describing where
price sat along the way, from a (lowest zone) up to the top of the
alphabet (highest zone). A sharp drop that goes quiet spells "ddaa". With
short words and a small alphabet, the market has a finite vocabulary:
256 words in the standard config, 6,561 in the long one. Every bar on
your chart is speaking exactly one of them.
The readout:
The current word translated into plain English:
"THE MARKET JUST SAID: sharp drop, now quiet at the lows" — plus how
many times this exact word has been seen on your chart and how rare
that is. A description of shape, not a signal.
The regime wash. The background tinted by the word's temperature —
toward red when the word lives in the low zones, toward green in the
high zones, stronger when the shape is more extreme.
The letter ribbon. A letter-colored track under price showing the
letter each bar is writing right now. Calm markets babble; you will see
the tone shift.
The Lexicon Panel
The full dictionary entry: seen-count, rarity
percentile, and causal win rate / average forward return over the next
1, 4 and 16 bars, computed on your chart's history with nothing peeking
into the future — each line carrying its validation verdict. Faint
numbers mean thin data, on purpose: your eyes should trust a statistic
exactly as much as the sample deserves.
Events. When the chart prints a word it has NEVER printed before,
you get a violet flare: "NEVER SEEN BEFORE". And on exactly two
market/timeframe combinations, a gold marker when a study-validated
word completes (more on both below — read the caveats).
Most words mean nothing
"no significant edge" is its most common verdict, by design.
Rare words looked like they preceded volatility in-sample on 5 of 7
markets. Out of sample: 0 of 28 configurations passed. The
"NEVER SEEN BEFORE" flare is a curiosity flag, and its own tooltip
tells you rarity's predictive power is not confirmed.
The vocabulary itself is remarkably stable: in 25 of 28 configurations
the word frequencies stayed the same from the research years into the
held-out years. Markets from 2008 and 2025 speak the same language.
It just mostly is not saying anything predictive.
Two words survived everything: "ddaa" on gold H1 (564 out-of-sample
occurrences; remained net-positive after median spread cost in the
study — informational, not a recommendation) and one BTC 1h pattern
that passed at 69% degradation against a 70% cutoff with 38 samples —
the panel labels that one "marginal" because that is what it is.
The surviving edges are breakpoint-fragile: they appear under the
empirically fitted letter boundaries, not under textbook Gaussian ones.
I tell you this because it is a real limitation.
What this tool is
A translator, not a signal. It tells you what the
current shape is called, how common it is, and what the record shows,
with the verdict attached to every number. Whether any of that adds
something to YOUR strategy is your test to run — that is not a
disclaimer, it is the design.
Frozen parameters
The letter boundaries, variance gates, and rarity
thresholds are hardcoded from the validation study (2026-07) and cannot
be edited or refit — there is deliberately no adaptive mode, because
refitting live would detach the display from its validation. Crypto
combos were validated on Binance USDT-M perpetual data; the indicator
accepts the same symbol on other venues since perp prices track closely
across exchanges — the panel tooltip names the exact validated
instrument. The panel's live statistics (seen-counts, win rates,
averages) are built from your chart's visible history; the validation
verdicts are not — they are the study's hardcoded results. On markets
outside the study, the indicator runs in EXPLORER MODE: fully functional,
clearly labeled, and it will never show you a validation star it did not
earn.
No Repainting:
Words finalize on bar close; statistics
update only when their forward horizons complete; nothing is drawn on
unconfirmed bars. Don't take my word for it: open bar replay, step
through a session, and try to catch a value changing after its bar
closes.
I wondered if it could. So I built a dictionary for the market. This
indicator is that dictionary — including everything it does NOT say.
How it reads the chart: Every bar, the recent closes are reduced to
their shape and spelled as a short word — a few letters describing where
price sat along the way, from a (lowest zone) up to the top of the
alphabet (highest zone). A sharp drop that goes quiet spells "ddaa". With
short words and a small alphabet, the market has a finite vocabulary:
256 words in the standard config, 6,561 in the long one. Every bar on
your chart is speaking exactly one of them.
The readout:
The current word translated into plain English:
"THE MARKET JUST SAID: sharp drop, now quiet at the lows" — plus how
many times this exact word has been seen on your chart and how rare
that is. A description of shape, not a signal.
The regime wash. The background tinted by the word's temperature —
toward red when the word lives in the low zones, toward green in the
high zones, stronger when the shape is more extreme.
The letter ribbon. A letter-colored track under price showing the
letter each bar is writing right now. Calm markets babble; you will see
the tone shift.
The Lexicon Panel
The full dictionary entry: seen-count, rarity
percentile, and causal win rate / average forward return over the next
1, 4 and 16 bars, computed on your chart's history with nothing peeking
into the future — each line carrying its validation verdict. Faint
numbers mean thin data, on purpose: your eyes should trust a statistic
exactly as much as the sample deserves.
Events. When the chart prints a word it has NEVER printed before,
you get a violet flare: "NEVER SEEN BEFORE". And on exactly two
market/timeframe combinations, a gold marker when a study-validated
word completes (more on both below — read the caveats).
Most words mean nothing
"no significant edge" is its most common verdict, by design.
Rare words looked like they preceded volatility in-sample on 5 of 7
markets. Out of sample: 0 of 28 configurations passed. The
"NEVER SEEN BEFORE" flare is a curiosity flag, and its own tooltip
tells you rarity's predictive power is not confirmed.
The vocabulary itself is remarkably stable: in 25 of 28 configurations
the word frequencies stayed the same from the research years into the
held-out years. Markets from 2008 and 2025 speak the same language.
It just mostly is not saying anything predictive.
Two words survived everything: "ddaa" on gold H1 (564 out-of-sample
occurrences; remained net-positive after median spread cost in the
study — informational, not a recommendation) and one BTC 1h pattern
that passed at 69% degradation against a 70% cutoff with 38 samples —
the panel labels that one "marginal" because that is what it is.
The surviving edges are breakpoint-fragile: they appear under the
empirically fitted letter boundaries, not under textbook Gaussian ones.
I tell you this because it is a real limitation.
What this tool is
A translator, not a signal. It tells you what the
current shape is called, how common it is, and what the record shows,
with the verdict attached to every number. Whether any of that adds
something to YOUR strategy is your test to run — that is not a
disclaimer, it is the design.
Frozen parameters
The letter boundaries, variance gates, and rarity
thresholds are hardcoded from the validation study (2026-07) and cannot
be edited or refit — there is deliberately no adaptive mode, because
refitting live would detach the display from its validation. Crypto
combos were validated on Binance USDT-M perpetual data; the indicator
accepts the same symbol on other venues since perp prices track closely
across exchanges — the panel tooltip names the exact validated
instrument. The panel's live statistics (seen-counts, win rates,
averages) are built from your chart's visible history; the validation
verdicts are not — they are the study's hardcoded results. On markets
outside the study, the indicator runs in EXPLORER MODE: fully functional,
clearly labeled, and it will never show you a validation star it did not
earn.
No Repainting:
Words finalize on bar close; statistics
update only when their forward horizons complete; nothing is drawn on
unconfirmed bars. Don't take my word for it: open bar replay, step
through a session, and try to catch a value changing after its bar
closes.
สคริปต์โอเพนซอร์ซ
ด้วยเจตนารมณ์หลักของ TradingView ผู้สร้างสคริปต์นี้ได้ทำให้เป็นโอเพนซอร์ส เพื่อให้เทรดเดอร์สามารถตรวจสอบและยืนยันฟังก์ชันการทำงานของมันได้ ขอชื่นชมผู้เขียน! แม้ว่าคุณจะใช้งานได้ฟรี แต่โปรดจำไว้ว่าการเผยแพร่โค้ดซ้ำจะต้องเป็นไปตาม กฎระเบียบการใช้งาน ของเรา
The Good, the Bad and the Bitcoin — No hype, just data.
Free indicators & honest backtests: youtube.com/@TheGoodTheBadAndTheBitcoin
Free indicators & honest backtests: youtube.com/@TheGoodTheBadAndTheBitcoin
คำจำกัดสิทธิ์ความรับผิดชอบ
ข้อมูลและบทความไม่ได้มีวัตถุประสงค์เพื่อก่อให้เกิดกิจกรรมทางการเงิน, การลงทุน, การซื้อขาย, ข้อเสนอแนะ หรือคำแนะนำประเภทอื่น ๆ ที่ให้หรือรับรองโดย TradingView อ่านเพิ่มเติมใน ข้อกำหนดการใช้งาน
สคริปต์โอเพนซอร์ซ
ด้วยเจตนารมณ์หลักของ TradingView ผู้สร้างสคริปต์นี้ได้ทำให้เป็นโอเพนซอร์ส เพื่อให้เทรดเดอร์สามารถตรวจสอบและยืนยันฟังก์ชันการทำงานของมันได้ ขอชื่นชมผู้เขียน! แม้ว่าคุณจะใช้งานได้ฟรี แต่โปรดจำไว้ว่าการเผยแพร่โค้ดซ้ำจะต้องเป็นไปตาม กฎระเบียบการใช้งาน ของเรา
The Good, the Bad and the Bitcoin — No hype, just data.
Free indicators & honest backtests: youtube.com/@TheGoodTheBadAndTheBitcoin
Free indicators & honest backtests: youtube.com/@TheGoodTheBadAndTheBitcoin
คำจำกัดสิทธิ์ความรับผิดชอบ
ข้อมูลและบทความไม่ได้มีวัตถุประสงค์เพื่อก่อให้เกิดกิจกรรมทางการเงิน, การลงทุน, การซื้อขาย, ข้อเสนอแนะ หรือคำแนะนำประเภทอื่น ๆ ที่ให้หรือรับรองโดย TradingView อ่านเพิ่มเติมใน ข้อกำหนดการใช้งาน