TickFoundry

Polymarket reads as an election venue from the outside. Over Aug 1-2 it traded $166M of notional. Esports was $55M of that - a third. Politics: $7.8M. League of Legends alone did $30.9M in two days, roughly 4x every political market combined.

New: TFVX, the crypto volatility Polymarket's prices imply over the next day (BTC, ETH, SOL, XRP, DOGE, BNB, HYPE and all coins). And the cost to move a typical market's odds by a quarter, by sector. From our order-book capture, updated daily. Free to view: tickfoundry.com/index

Looked for bid-ask bounce in Polymarket's trade tape. Mostly not there. Oct 18, 1.30M prints: of 536,451 successive nonzero price changes on the same token, 48.9% reversed the last move, 51.1% continued. Lag-1 correlation: -0.044. Roll-style bounce would put reversals well above 50%.

Barcelona-win market (Oct 17), YES token: one print of 126,341 shares at 99.9¢ = $126,215, at 18:18 UTC. The buyer's maximum gain: $126. All 425 prints at 99¢+ in that market that day: $260,877 notional, $515 of upside. The same token averaged 67¢ in the 14:00 UTC hour. Trade tape, one market.

Polymarket's tape is never quiet. On Oct 16, 96.7% of the 86,400 seconds had at least one print; the longest gap was 4 seconds. Median second: 5 prints. Busiest: 140 (23:35:46 UTC). Dollars are lumpier: the top 1% of seconds by value carried 35.2% of the $38.6M. Trade tape, one day.

One print size dominates Polymarket's tape. On Oct 15, exactly 5 shares was 10.8% of 595,141 prints. Next: 10 shares (4.4%), 20 (2.5%). 40.3% of prints were whole-share sizes, but only 29.3% of the $49.9M. Prints of 1,000+ shares: 2.5% of prints, 65.1% of dollars. One day, trade tape.

Coding agents as compilers is a good analogy. When I started, there were devs that insisted anyone using a high level language like C was basically a vibe coding noob generating garbage assembly. Those devs would've been much better off finding cases where their deep assembly knowledge did matter.

Polymarket's active universe turns over fast. Of 25,352 tokens that traded on Oct 14, 63.9% had no print on any of the six days before (Oct 8-13). They carried 22.8% of the $51.7M. Only 28.5% of Oct 13's tokens traded again the next day. A week-old token list misses most of today's. Trade tape.

Price change between consecutive prints, same Polymarket token, Oct 13: 44.8% of 569,665 pairs traded at the identical price. 70.0% moved 1¢ or less, 5.5% jumped 10¢ or more, median move 0.6¢. Dollar-weighted, 57.4% of volume printed at an unchanged price. One day of trade tape.

Polymarket's side label is lopsided, and the answer depends on how you count. On Oct 12, 87.2% of 563,241 prints were labelled BUY, but only 78.4% of the $40.6M. Above 90¢: 80.7% of prints, 59.2% of dollars. SELL prints are fewer and bigger ($121 avg vs $65). Trade tape, one day.

Polymarket volume is extremely concentrated. On Oct 11, 30,118 tokens traded. The top 1% (302 tokens) did 64.9% of the $49.5M. 117 tokens made half. Meanwhile 60.2% of tokens traded under $100 all day, together 0.9% of dollars. Averages across "all markets" mostly describe the tail. Trade tape.

Share counts mislead on cheap contracts. On Oct 10, Polymarket prints under 10¢ were 6.8% of 747,049 prints and 12.5% of shares, but 0.8% of dollars. Median print there: 40 shares, $1.20. At 90¢+: 19.6% of prints, 34.9% of dollars. Weight by price×size. Measured on the trade tape.

Art Institute of Chicago's collection data is unusually reusable: REST API plus nightly JSON dumps for artworks, artists, places and exhibitions, with IIIF image metadata. Useful for cultural heritage search/CV. Idea: map metadata gaps by decade and medium. https://api.artic.edu/docs/ #OpenData

🐋 Whale Trade Detected 📊 Israel x Iran ceasefire continues through October 15? 💰 $6,700 BUY Yes @ 96c 🔴 Score: 94/100 👤 Trader: Dental-Liquid | 733 days old https://polymarket.com/profile/0x7f9e2d1df78614564a70becc7fa14aa9a6623a0e

Polymarket by UTC hour, Oct 9. The busiest hour for prints (17:00) had 2.4x the prints of the quietest (00:00). In dollars the gap was 8.4x: $3.70M at 16:00 vs $0.44M at 01:00. Quiet hours aren't just fewer trades, they're smaller ones. One day, trade tape.

Where Polymarket prints land vs. the book. On Oct 8, of 546,603 prints with a prior top-of-book quote, 76.1% were exactly at the best bid or ask, 20.2% inside the spread, 3.7% outside the last quote on file. Benchmarked to mid, most prints just measure half the spread. Trade tape + L1.

Boring result: on Oct 7, Polymarket's trade tape had 621,459 prints and 621,459 distinct transaction hashes. No hash appeared twice, across 21,278 tokens. Grouping or deduping prints by hash returns the print count, not matches or orders. Measured on the trade tape.

Polymarket trades arrive in bursts. On Oct 6, 30.1% of 600,588 prints came within 1 second of the previous print on the same token, 5.9% within 10 ms. The median gap on a token was 5.0 s. Models that treat prints as independent arrivals will understate clustering. Measured on the trade tape.

Polymarket prices aren't on a 1¢ grid everywhere. On Oct 5, 13.3% of 589,855 trade prints, carrying 26.2% of the $47.6M notional, were priced off a whole cent (0.048, 0.003). Under 5¢, 61% of prints were off-cent; over 95¢, 52%. Rounding to cents distorts the tails. Measured on the trade tape.

Polymarket print sizes cluster hard. On Oct 4, 8.4% of 692,555 prints were exactly 5 shares, nearly 2x the next most common size (10 shares, 4.4%). 6.5% were worth $1.00 to the cent. Whole-share sizes were 38% of prints but 48% of notional. Measured on the trade tape.

Polymarket's trade tape, Oct 3: the median print was $6.08, and 62% of prints were under $10. Together they were 3.0% of the $56.9M notional. The top 1% of prints (each $1,130+) carried 52%. A typical trade and a typical dollar are very different things. Measured on the trade tape.

Averaged over every Polymarket token with a live two-sided quote on Oct 2, the median spread was 84¢. Meaningless: 87% of those 185k tokens had no trade that day. The 23k that did had a median spread of 2.3¢, and 43% were 2¢ or tighter. Time-weighted, tokens with 1h+ of quotes.

Yesterday on Polymarket's trade tape, 87.1% of prints were labelled BUY, carrying 78.7% of the $47.4M notional. SELL prints were rarer but bigger: $127 on average vs $70. The label is per token, and every market has two, so the BUY/SELL split is not net buying pressure.

Yesterday on Polymarket, 18:00 UTC carried $4.38M of trade notional and 01:00 UTC $0.42M, about 10x apart. Trade counts differed only 2.4x (38,763 vs 16,354): the average print grew from $26 to $113. Sizing liquidity by trade count understates busy hours. Measured on the trade tape.

Yesterday 30% of Polymarket's $35.1M trade notional printed at 95¢ or higher, on 17% of trades. Only 0.4% printed at 5¢ or below. Volume counts $99 bought at 99¢ the same as $99 at 50¢, though the first has about 1¢ of upside per share. Measured on the trade tape, by token price.

Polymarket volume is more lopsided than "top markets" lists suggest. Of 14,267 markets that traded yesterday, the top 1% (142) took 63% of the $28.9M notional, the top 10% took 92%, and the median market traded $65. 57% never reached $100. Measured on the trade tape, by condition ID.

Biggest single trade yesterday on Polymarket's esports markets: $72,864 bought on Natus Vincere to win Game 1 vs Aurora (Dota 2) — one fill of 72,937 contracts at 99.9¢, 9.9% of that market's entire day volume. Order book: tickfoundry.com

On yesterday's UEFA Nations League markets, 13% of dollar volume traded after price had already hit 98%+ (or under 2%) — after the outcome was basically locked in. Why keep trading there? Position unwinds, incentive farming, bots that don't check price? Genuinely unsure which dominates.

New: TickFoundry Lab. Replay any Polymarket or Hyperliquid order book tick by tick in your browser, and walk an order of any size through the recorded ladder to see its real fill and slippage. Included with every account. The sample day opens with no account: tickfoundry.com/lab/replay

Polymarket perps quotes lead Hyperliquid by 250–500 ms on source clocks: 43 of 44 same-symbol pairs over 42 days, and it held after we fixed a bug in our own HL book data. Source time, not a tradable latency edge. Full method and tables at the link.

Method note: a market's trade tape undercounts its real activity. Yesterday Polymarket's ETH Up-or-Down 5-minute markets saw 69,567 trades, but their order books updated 7,165,589 times: about 103 quote changes for every trade that printed.

Biggest single trade yesterday on Polymarket's Ethereum price-target markets: $80,071 bought on NO for 'Will Ethereum dip to $2,250 by Dec 31, 2026?' — one fill of 106,761 contracts at 75¢. Order book: tickfoundry.com

Tested whether Polymarket trade sizes cluster at round numbers, like retail sports bets usually do (10, 50, 100 units). Yesterday's NFL slate: 56,921 trades, $32.8M notional. 55.6% of sizes are non-integer; only 6.5% land on an exact multiple of 100. Most volume isn't round-number sized.

Myth-check: two order books (YES + NO tokens) shouldn't let you lock in a free profit. Yesterday's Houston-Texas Tech book on Polymarket mostly held: buying both sides averaged 101.4¢ across 27k paired quotes. But 25 times it inverted, once to 96¢ combined — a real, if brief, riskless 4¢.

Yesterday's Polymarket book on "Will Chelsea win?" (Brentford v Chelsea) held a 1¢ spread 99.4% of a 20.7-hour session: 40¢ pre-match, drifting to 34¢ by kickoff, then under a cent within 45 minutes as trading resolved it. Order book: tickfoundry.com

Same platform, same day, wildly different trader behavior. Yesterday's average trade size on Polymarket: NFL markets $1,087/trade ($6.13M / 5,645 trades), BTC 5-min up/down markets $15/trade ($5.12M / 335,242 trades). A 71x gap hiding inside similar dollar volume.

Polymarket's odds on a 25bp Fed hike this month: 78.5% Sunday → 88.5% Tuesday. The complementary "no change" market moved the same 10 points the other way, 20.5% → 10.5%. $3.74M traded across the two Tuesday alone. Order book: tickfoundry.com

Yesterday's biggest single trade on Polymarket: a $1.50M sell print hit 'Will AS Roma win on 2026-09-14?' at 99.9¢. That's 53% of that market's entire day, and 5.8x the next-largest trade on the whole platform. Full order book: tickfoundry.com

Yesterday's biggest esports market (League of Legends, $13.06M across 50,878 trades): the top 1% of trades by size made up 45% of the dollar volume, median trade was just $18. Is that concentration typical of prediction markets, or specific to fast live-betting books?

Biggest single trade on Polymarket yesterday: $359.6K bought on Rayo Vallecano NOT covering a -1.5 spread vs Real Madrid, at 99.2¢. Unlike most whale prints, the price didn't move: the book had already sat near 99% for hours before this trade landed. Order book: tickfoundry.com

Tested an assumption: order books should tighten during US trading hours, when more traders are online. Checked Polymarket's BTC 15-min up-or-down books yesterday, median bid-ask spread by hour (UTC): 1¢ in all 24 hours, no exception, across 4.3M quotes. 89.5% sat at that floor, day and night.

Yesterday's most-traded prediction market wasn't the tightest one. League of Legends match books: 23,355 trades, 29¢ median bid-ask spread. BTC 15m up-or-down: 105,265 trades, 1¢ spread. Trade count alone doesn't predict spread; market structure does.

Biggest single trade on Polymarket yesterday: $142K bought pregame into the Dodgers side of Nationals–Dodgers at 64¢. That price fell as low as 23¢ mid-game before climbing to 99.9¢ by the finish. Full order book: tickfoundry.com

Open question: yesterday, 21% of all dollar volume on Polymarket ($14.4M of $68.3M) traded at prices above 98c or below 2c, on markets already close to decided. League of Legends alone did $5.0M of its $15.6M there. Why does so much money move once the outcome is barely in doubt?

Book stat: yesterday's biggest La Liga match on Polymarket, Real Madrid to win, traded $921.7K across 4,582 fills. The book stayed tight throughout: time-weighted spread of 1.01¢ (median 1¢) over 21 hours of quotes. Full L1/L2 tick data: tickfoundry.com

Trade count and dollar volume measure different things. Yesterday on Polymarket, ETH Up-or-Down 5m logged the most trades of any market: 92,428, nearly 3x League of Legends' 34,856. LoL still did 13x the dollar volume ($8.5M vs $639K). Average trade size: $243.86 (LoL) vs $6.91 (ETH-5m).

Biggest print yesterday: $218K sold into the ATP Aguilar–Rublev match at 99.9¢, as the book swung from Aguilar 38% to 99.9% in about 3 hours (15:02→18:11 UTC). Rublev had been favored most of the match. Full L1/L2 tick data: tickfoundry.com

Odds move: Polymarket's Fed 25bp-cut market (Sept 2026 meeting) moved 54.5%→58.5% implied over the last 24h; the 'no change' market fell 43.5%→40.5%. $649K traded across the two yesterday. This is the live order-book mid, not the settled trade tape. Full L1/L2 tick data: tickfoundry.com

Method note: checked every trade in a 15-minute BTC market yesterday against the book at that instant. ~70% printed exactly at the best bid or best ask, not between them (69.7% and 72.6% on the two sides, 1,797 and 1,879 trades). The tape mostly shows the edge of the book, not its middle.

Tested a natural assumption: a big trade should spike book activity right after it prints. Checked two $1.3K+ trades in yesterday's BTC 5-min markets: no spike either time. Update rates were already 500-1,400 updates/sec before each trade, then kept falling after.

Yesterday's biggest Premier League book: 'Will Liverpool win?' ($1.4M traded, 5,632 fills). Over a wild 102-minute stretch the YES price swung 63.5¢→1.5¢, but the book held a ≤1¢ spread 92% of that time (1.10¢ time-weighted). Full L1/L2 tick data: tickfoundry.com