How to Tell Time
with BitClock

What this manual is built from

Three sources, and nothing else.

The BitClock archive. 21,281,769 order-book snapshots of the BTC-USD book, recorded roughly once per second from 2025-10-12 20:25 UTC to 2026-07-07 12:27 UTC — 269 days. Every statistic in this manual was recomputed directly from this archive. From 2026-07-07 the production collector — the same feed the dashboard serves its customers — continues the record without interruption; rows past the freeze are marked ‡ and were recomputed directly from it on 2026-08-21. The combined record passed 26 million snapshots in late August 2026 and grows every second. One dated method change: through 2026-08-21 every reading is a once-per-second snapshot (what the dial showed); from 2026-08-22 the collector captures each second’s true high and low across every book state, and the product’s extremes display them. The appendix states each computation’s method.

The designer’s own account. A recorded interview from 2026-08-08 covering how the instrument was invented and how he reads it. Where his words carry the meaning, they are quoted verbatim.

The product’s source code. The formula, the band, the two display faces, and the daily extremes are stated exactly as the shipping code computes them.

Contents

The one number

BitClock reads the live BTC-USD order book — the standing list of resting buy orders (bids) and resting sell orders (asks). Once per second it does one piece of arithmetic.

A ratio of 1 means the two sides inside the band are equal. Above 1, the ask side is bigger; below 1, the bid side is bigger. At the moment on the cover, 1.91 BTC of asks stood against 330.58 BTC of bids: 1.91 ÷ 330.58 = 0.006.

Two design decisions in that formula are the designer’s, and he explained both. The centre line at 1 was a deliberate goal — he tried the arithmetic “a few different ways” and kept the version built around 1 as the balance point, because that is the system he wanted to read. And the ±$2,400 band is inherited, not optimized:

“That’s the bandwidth that’s presented in the depth-of-market chart that I was watching and learned this stuff from. So that is why I specifically picked that.”— the designer, on why the band is ±$2,400

The dashboard shows the two depth totals alongside the ratio rather than the ratio alone. That, too, is deliberate: the same ratio on a thin book and a thick book is not the same reading to him — in his words, it matters whether “we’re at 900/100, or we’re at 90/10.”

R-HOD and R-LOD

BitClock also tracks the highest and lowest ratio printed so far each day — R-HOD and R-LOD — each stored with the BTC price at the moment the extreme printed. The reason these exist is one of this manual’s central facts: the extremes last seconds, and a person watching by hand does not catch them.

“It won’t hold at the high of the day. It’ll flash the high of the day, and it won’t hold at the low of the day.”

Watching by hand for months, the designer systematically never saw the widest readings — they were too brief — and only discovered their true width when the software began logging every second. (From 2026-08-22, R-HOD and R-LOD are each second’s true high and low across every full book state — they can run beyond anything the dial itself displayed, because the live number is still a once-per-second sample. In the first hour measured, the true extreme ran beyond the sample in 92% of seconds — usually by a hair, once by 2.31 inside a single second.) His use of the extremes is retrospective within the day: “if they already stuck it, it gives you an idea if the thing’s over or not.”

The two faces of the clock

Two user-selectable display scales — same data, two dials. This manual states everything on the Raw face unless marked otherwise.

Where the number came from

Before BitClock was software, it was a pen.

For roughly four months the designer hand-logged the book — college-ruled paper, one line per reading, ten to fifteen seconds to write, BTC-USD only. Each line: date and time, price, the contract count on each side, how far price had moved since the last line — and, once the method matured, the ratio between the sides worked out by hand, and how far it had shifted. His own example of the format: “86,234 dollars on February 2nd at 10 o’clock A.M. with 800 on the bid and 300 on the ask, giving a ratio of this, and price moved this much from my last note.” Pages and pages of those lines.

He logged at the desk, multiple times a day, at dull moments and sharp ones. Three things always prompted a reading: sitting down at the desk for any reason, any aggressive move in the market, and the moments his existing tools fired — when the TDI was tapping, and when the 50, 200 or 800-period moving averages got tapped. The question each time was the same: where is the ratio, and how are the contracts stacked, at this exact moment?

Out of those months he stopped classifying ticks and started classifying days: grind-up days (a small new high, a pullback, sideways chop, another small new high, for 8–14 hours), their mirror on the way down, fakeout and stop-hunt days carrying extreme disparity, and accumulation days. His reasoning for why the day is the right unit, verbatim: “they kind of got to hold a particular stack all day if they’re gonna move it in a particular way all day.”

The formula was born from friction, not a eureka. After six to eight weeks the loop — open the exchange, find the depth chart, read the numbers, punch the calculator, write the note — had become intolerable:

“I realized I needed a way to summarize this information into a much more digestible fashion.”

So he built one number with 1 as its centre line, then a first dashboard — terminal-based, then HTML, assembled with AI assistance and, in his description, “a lot of copy-paste and very little understanding and a steep learning curve.” Hand-logging ended the day the software started:

“Hand logging never scaled. It always sucked, but that was the only way I knew how to do it. I had to start there.”

One more thing pushed him to commit to it. He could not back-test the idea, because no broker keeps this data — and he read that absence as information:

“It was using data that none of the brokers would allow you to view historically, so there’s no way I could look back and say is this valid — which that in itself for me was a big clue that it was in fact valid. They’ll let you look at everything under the sun, and the fact that they don’t cache this data, when it’s part of every order feed, just seemed obtuse.”

That is why the archive underneath this manual exists: BitClock has been logging the number every second since October 2025, precisely because nobody else does.

The nine months on record

Everything measured in this manual comes from one specific stretch of market: a bear market with three sharp rallies inside it. Prices are daily closes of the archive’s own mid-price record.

Swing turns from a 10% zigzag on daily closes. The archive’s final close is 63,672 on 2026-07-07 — 45% below its first day; the June 30 low sat 49% below it.

Two feed failures are excluded from every computation that follows: the whole of 2026-05-08 (the collector recorded garbage for most of that day, including 3,586 spurious sub-0.1 prints and mid-prices near half the true price) and a ten-minute window on 2026-06-09 (whose spike printed a ratio of 19,641). A further handful of one-to-two-second “book teleport” errors — where one side’s recorded depth collapses to a fraction of its surrounding values for a single tick and instantly snaps back — were identified by inspecting the seconds around every extreme print; the appendix lists each one excluded, with the screening rule. Out of 21.3 million rows, exactly 4 recorded a ratio of zero.

What the dial did: the five brackets

Slicing the cleaned archive into five-minute intervals and bracketing each by its average ratio gives the base map of the instrument.

Forward returns are measured on the archive’s own mid-price 24 and 72 hours ahead. The up-leg / down-leg split assigns each interval to the rally or decline phase it sat in (Chapter 4’s swing map).

Counts in parentheses. May 8 and the June 9 window excluded. Historical frequencies of one nine-month bear market — not probabilities, and not predictions.

Row by row: the rare sub-0.2 intervals were the only bracket with positive forward returns in both rally and decline phases. The broad 0.2–0.8 bracket meant opposite things depending on the phase — +2.1% over three days inside rallies, −2.0% inside declines. The majority of the tape, 0.8–2.25, carried no measurable edge in either direction. From 2.25 up, forward returns were negative even inside rallies. And the 27 intervals at 5.00 or higher — spread over just four days — carried the worst forward returns on the dial: −2.9% averaged over the next 24 hours, −3.7% over three days.

The finer slices

Below 0.2, depth mattered too: the 0.1–0.2 slice is where the 3-day numbers turned positive (+1.55%); the deeper sub-0.1 prints are too rare to slice further (46 intervals on 3 days). Inside 0.2–0.8, depth mattered: within rallies, 0.2–0.5 returned +3.06% over 3 days against +1.71% for 0.5–0.8. 2.0–2.25 was a stall, not a signal (+0.02% over 3 days inside rallies); the negative numbers start at 2.25.

The ceiling

Out of 21,281,769 snapshots, exactly 1,336 printed a ratio of 7.0 or higher — 0.006% of the tape, across 15 days. After removing the feed errors (appendix), nine real episodes on eight days remain; the production collector has added 22 more on nine days — four once-per-second (‡), then 18 true-extremes from the evening of August 21 (§).

Forward returns from that day’s close. The only sustained multi-minute runs above 7 in the whole record — October 20, 26 and 28 — all sit in the final two weeks before the market rolled over into the November decline.

The designer’s recollection, given before this census was run, was that “above roughly 9 or 10, you top out.” The archive agreed — its three highest peaks were 11.13, 10.65 and 9.54 — but the live era has run past that ceiling repeatedly: nine 10+ episodes on six days through August 28, topping at 15.05, 13.51§ and 12.94§ (windows still open). In the settled record, the October 28 double-run, January 13 and May 6 prints each came one to four days before a major swing high, with slides of 10–35% to follow. The record is not unanimous — December 18 was followed by a bounce — and where the archive produced eight extreme days in nine months, the production era added nine in ten days (August 19–28). Part of that cadence is the stricter true-extremes measurement (§), part the hot regime — a method seam, not a market transformed overnight.

The record book: ten heaviest days

Each day’s single highest reading, with the book that produced it — both faces of the dial, the resting contracts on each side, and what price did next.

Forward moves measured from the tick’s own timestamp on the cleaned mid-price series. “30d max” is the largest excursion from the print’s price within 30 days; “R pre-max” is the most extreme five-minute-average ratio in the 72 hours leading into that price extreme (minimum into a bottom, maximum into a top) — the book signal, where it precedes the final price extension. Confirmed feed-teleport days (appendix) excluded. ‡/§Production-collector rows, recorded after the archive froze — the same feed the dashboard serves; 30-day windows still open as of 2026-08-29 (—/†), and “R pre-max” is computed once a row’s window closes. From 2026-08-22 (in practice the evening of August 21), entries in either record book are each second’s true high or low (full-book states only), no longer one-per-second snapshots — a stricter record that can only run more extreme; § marks those window-sourced rows. A § row’s Bid/Ask cells are its second’s once-per-second snapshot (the collector stores only the window extreme’s ratio), so ask ÷ bid there reproduces the sampled second — 7.53, 7.69, 10.71 and 12.03 on the four § rows — not the peak. The August 19 row’s previously printed +6.5% could not be re-derived and is corrected to +4.1%.

What the ten rows share: an ask wall of 118–1,245 BTC standing over a bid side of 8–130 BTC. The five live rows’ 30-day windows are all still open; among the five settled, archive-era rows, four were lower seven days on — the exception, October 20 (+4.0%), still fell 2.9% within the first 24 hours. A month out, four of the five settled rows were still lower, two of them by more than 20%. And the last two columns close the loop on the configuration idea: the four whose biggest 30-day move was a decline all saw the book drain to 0.12–0.36 inside the final three days into the bottom, while December 18 — the one that resolved upward — saw it stack to 4.67 into the top — the far end of the move carried the opposite reading, and that signal led the price extreme itself by as much as 37 hours.

The five live-era rows are production-feed records — the same numbers the dashboard and apps served their viewers — and their screening is stronger than anything the legacy archive allowed: the collector records how many price levels each side of the book is built from, and the bid side carried 20,056 levels at the 15.052 print, where a feed rebuild shows a few hundred. August 19 is the new record ceiling: 15.052 printed at the crest of a violent +3.8% rally on a bid band drained to 7.85 BTC, and its price has run +17.3% in the days since — straight through the stretch whose § peaks fill the next four rows, the stacked-book readings arriving at the far end of the move exactly as the settled rows describe. All five windows are still open; whether they mark a top the way October and January did, the next month will tell.

The floor

17,609 snapshots printed below 0.1 — 0.08% of the tape. Excluding the May 8 outage and days whose only sub-0.1 prints were one or two isolated ticks, twelve days carried real sub-0.1 runs, arriving in four clusters. The deepest sustained floor is November 14’s 0.0377 — below 0.1 for stretches of up to nineteen unbroken minutes; the record floor of November 21 was a nineteen-second flash.

November 14 stands alone: sub-0.1 prints recurring across an eighteen-hour stretch, totalling three and a half hours below 0.1 — by tick count, more than every other day in the archive combined.

The pattern the dates draw: in each cluster, the first extreme days were followed by more decline (November 13: −13.1% over the next week; January 30: −16.1%), while the last extreme day of a cluster landed at or within days of a swing turn (November 21, February 6, June 25). Which print is the last is only knowable afterwards — that is a limit of the record, not a technique.

This floor is also the one region the designer described numerically, unprompted, when asked how he reads the book on a stop hunt: the ratio runs to roughly 0.05–0.09 accompanied by large candles, and those candles then get a retest that lines up with his TDI entry criteria on the 4-hour and daily timeframes. The archive’s twelve real sub-0.1 days all sat inside capitulation stretches, consistent with his description; his full stop-hunt read is his method, described in Chapter 10, not a measured claim.

The record book: ten emptiest days

Each day’s single lowest tick, with the book that produced it. Note the shape every real floor print shares: a large-to-huge bid stack against a sell side that collapsed. The record row was a nineteen-second flash sweep; the deepest sustained floor belongs to November 14.

Forward moves measured from the tick’s own timestamp on the cleaned five-minute mid-price series. The June 24 and 25 windows, open when this manual was first compiled, were completed from the production record: both resolved upward, topping +11.0% and +13.6% on July 21 into a book stacked to 2.96, the signal leading that top by 38 hours. “30d max” and “R pre-max” as defined in Chapter 6. Two former top-10 candidates — 2026-02-24 (0.0269) and 2025-12-02 (0.0710) — were re-inspected for this table and confirmed as one-tick feed teleports; they are excluded (appendix).

Every row is the same shape: a bid stack of 330 to 1,166 BTC against an ask side of 1.9 to 76 BTC. The forward record splits by position in the cluster, exactly as Chapter 7’s cluster table shows: prints that landed early in a capitulation stretch still had the worst of it ahead (November 14: −14.3% the following week; January 30: −20.7%; November 16: −7.6%), while prints that landed at a cluster’s end sat days from a turn (November 21: +9.0%, swing low next day; February 6: +15.9% in 24 hours, one day after the February 5 swing low; June 25: +5.2%, five days before the June 30 low).

Which kind of print you are looking at is only knowable afterwards. That asymmetry — obvious in the table, invisible at the moment of the print — is the single most important honest fact about the floor. The last two columns repeat Chapter 6’s mirror pattern from the other side: every recovery that followed a floor print topped out into a stacked book (1.85–4.36), and every continued decline bottomed into a drained one (0.14–0.23), with the book signal leading the price extreme by up to 39 hours.

The daily view and the trend

Averaging the ratio per day and placing it beside the swing map produces the cleanest single set of facts in the archive.

At the six swing turns

Every swing high printed a day-ratio between 1.39 and 2.31; every swing low between 0.45 and 0.98. The three rallies each began with a light book — the first five days of the three up-legs averaged 1.01, 0.95 and 0.83.

The same reading in different trends

Same instrument, opposite outcomes by phase. The samples are small — single digits in three rows — and are stated so you can weigh them accordingly.

The five heaviest days by daily average

All five were followed by a lower price three days on. The two heaviest — November 1 and 2 — sat at the last local high before the November decline; the fourth is the January 14 swing high itself.

The other hand of the clock: the TDI

The designer does not read the ratio alone. His stated prerequisite for anyone learning his approach: “I’d want them to understand the market maker method and price action.”

The indicator that pairs with the ratio in his process is the Traders Dynamic Index — created by Dean Malone in the early 2000s and freely available in most charting platforms — read the way the Beat the Market Maker method (Steve Mauro) teaches it. Both the indicator and the method are third-party material; what follows states his usage, and then what this archive’s price record shows when his rule is replayed on it.

The TDI as he uses it: an RSI-based panel where a fast green price line moves inside a volatility envelope (the “water”) drawn around a slow base line, with marked extremes at 68 and 32. The event that matters is the shark fin: the green line leaves the water, reaches the extreme, and then closes back inside — the return into the water is the signal (“the lock”), not the excursion itself.

His rule, replayed on the archive’s daily closes

Applying that definition mechanically — green line exits the envelope, touches 68 (sell) or 32 (buy), signal on the close back inside — produces exactly three strict signals in nine months:

Loosening the extremes to 65/35 adds two sell fins that went nowhere (Jan 8: +2.9% in 10 days; Apr 18: +0.8%) and one buy fin that failed badly (May 31: −16.5% in 10 days). The strict 68/32 rule is what filtered those out.

The one strict sell fin caught the largest decline of the year. The two strict buy fins marked places where the fall paused — neither produced a sustained rally, which is consistent with everything else in this bear-market record: nothing made longs reliably profitable inside declines.

What the book added to the TDI

“Oversold” alone kept falling. The 15 oversold days that also carried a drained book were the only cohort that did not — though on this tape they merely stopped the bleeding rather than launched rallies.

How the designer actually reads it

His method in his own words, from the 2026-08-08 interview — described, not prescribed.

Why it lives on a lock screen

“That’s the whole point of it. It’s a widget that lives on top of the lock screen. So every time I tap my phone, I see it… It’s like a clock on the wall.”

“There’s a bunch of time when nothing’s happening and nothing’s going on and it’ll eat your soul if you let it. So this just makes it convenient to see when it’s important to start to pay attention.”

The archive agrees with the premise: 57.3% of the tape sat in the 0.8–2.25 bracket that carried no measurable edge. Most of the time, the honest reading is that there is nothing to read.

On systems and early luck — his one piece of general advice to a newcomer, verbatim:

“Honestly the worst thing that can happen is you make good money on a system that sucks due to dumb luck right out of the gate, because you think it’s then something that it’s not. And then you’re trying to chase this thing that just worked the first time, and you can get yourself into trouble.”

Honest limits

Six things this manual cannot claim, stated as plainly as everything it does.

One market, one regime. Every number here was measured on nine months of a bear market: a 45% net decline containing three rallies of 14.5–24.5%. No bull-market data exists in this archive. Whether any bracket behaves the same way in a different regime is unmeasured.

Frequencies, not probabilities. “Sub-0.2 intervals averaged +0.92% over the next day” is a fact about the past tape. It is not a forecast, and several of the strongest-looking rows rest on single-digit day counts, which are printed beside them for exactly that reason.

The designer’s trade history is not yet correlated. He deferred every question about specific historical episodes until his personal trading record can be laid against this archive — his words: “we will leave those open for now.” Until that work exists, this manual contains no claim about what he traded or earned, and his level recollections are presented beside the measured record, not as verified fact.

The TDI and the market maker method are not BitClock’s. The Traders Dynamic Index is the work of Dean Malone; the Beat the Market Maker method is the work of Steve Mauro. Both are third-party material, described here only as the designer’s usage of them.

The dial is silent most of the time. The no-edge bracket covered 57.3% of the tape, and the two decisive extremes — sustained 5+ and sub-0.1 runs — occurred on four and twelve days respectively, out of 269.

This is educational material, not financial advice. Nothing here is a recommendation to buy or sell anything. Trading cryptocurrency involves substantial risk of loss; position sizes should assume that any read, including these, can be wrong.

Verification

Every figure in this manual was recomputed from the source archive (21,281,769 rows, read-only). The methods, by claim family:

Excluded data, in full

Feed outages: all of 2026-05-08 (garbage prices and 3,586 spurious sub-0.1 prints), and 2026-06-09 22:35–22:45 UTC (spike to 19,641). Recording gap: 2026-08-27 10:51–15:32 UTC (4 h 41 m, hosting outage) — nothing recorded or displayed; that day’s extremes are not comparable to full days. One-tick book-teleport errors excluded from the ceiling census, identified by depth collapsing implausibly for 1–2 seconds against its surroundings: 2025-11-19 (14.73), 2025-11-20 (7.09 — the recorded bid side fell 415→10 BTC for exactly two ticks and snapped back), 2025-12-05 (21.62), 2026-02-17 (76.82), 2026-02-19 (47.50), 2026-04-16 (14.70). Isolated 1–2-tick sub-0.1 prints excluded from the floor census: 2025-11-07, 2026-02-01, 2026-06-26. Two further isolated days, re-inspected 2026-08-09, confirmed as one-tick book teleports: 2025-12-02 (recorded ask side fell 250→16 BTC for a single tick inside a tape trading steadily at 1.1) and 2026-02-24 (both sides collapsed for a single tick against a steady 0.40 book). Four rows in the archive record a ratio of exactly zero (three days; both depths zero — empty feed messages). In the production collector’s record (after 2026-07-07), book-rebuild artifacts are excluded by its level-count screen (Ch. 6); the only one that would otherwise reach a table is 2026-08-13 (“0.096” on 106 bid levels). Nothing else was excluded.

Compiled 2026-08-08; revised 2026-08-09, 2026-08-20, 2026-08-21, 2026-08-29 (records §; Normalized scale → Option B), 2026-09-02. Sources: the BitClock archive (21,281,769 snapshots, 2025-10-12 → 2026-07-07 UTC), the production collector’s live feed, the designer’s 2026-08-08 interview, and the product’s source code. The Traders Dynamic Index is Dean Malone’s; the Beat the Market Maker method is Steve Mauro’s; references describe the designer’s personal usage of both. All statistics are historical frequencies from one nine-month bear market plus the marked live-feed rows. Educational material only — not investment advice.