← readthetape.cc Chart-reading guides Monkey Index
Read the Tape · the daily chart game

The Tape Report

No. 2 · the first monthJuly 5 to August 1, 2026

Five real S&P 500 charts a day, tickers and dates hidden, everyone gets the same hand. Players call the next five trading days and a low, medium or high confidence pick sizes the bet. Across 30,353 calls in month one, players raised confidence after losses and trimmed it after wins.

Read the Tape is a free daily game where players view five historical S&P 500 price charts with tickers and dates hidden and all prices normalised to a starting value of 100. This report covers the first month of active game play, from July 5 to August 1, 2026.

Across 28 daily sessions, active players evaluated 140 charts and submitted 30,353 calls. The mean stated confidence was 74.0%, and the overall hit rate was 54.3%. Because every active participant plays the same five charts daily, these calls are not independent observations. The effective sample size is 140 charts rather than 30,353 calls, so one heavy day can swing a week of numbers. During fortnight one, players hit 53.8%; the second fortnight, with four times the volume at 24,339 calls, hit 54.5%. The early numbers held.

the flat curve

The stated confidence level carries almost no information regarding whether a call lands. Roughly 30 extra points of stated confidence buy about 4 points of accuracy. The low bracket rose from 47.4% in the fortnight report, so the curve is now flatter than it was.

perfect calibration low · 55-64% hit 52.1% medium · 65-74% hit 54.1% high · 85-94% hit 56.0% said 90%, hit 56%
Realised hit rate by stated confidence bracket, all calls July 5 to August 1. Brackets 75-84% and 95-100% held 13 calls between them and are omitted.
stated confidencecallshit rate
low · 55-64%5,35052.1%
medium · 65-74%14,96954.1%
high · 85-94%10,02156.0%
the double down

Across 23,222 consecutive same-session call pairs, players raised their stated confidence by an average of 3.5 points after a losing call, and trimmed it by 0.75 points after a win. Mean next-call confidence stood at 76.5% following a loss, compared to 73.1% following a win. Only 6.7% of players cut confidence after a wrong call, while 19.8% cut it after a right one. The modal response to being wrong was to press harder. The modal response to being right was to take chips off. The decline after wins rules out simple intra-session confidence drift. The pattern is the exact inverse of calibrated updating.

cut confidence after a win 19.8% cut confidence after a loss 6.7%
Only 6.7% of players cut stated confidence after a losing call, compared to 19.8% after a win.
the bear inside

A bearish bias remained visible in the data. Down calls accounted for 41.4% of all submissions (12,556 calls) and hit 43.5%. Bullish up calls (17,797 calls) hit a 62.0% success rate. Over the month, 81 of 140 charts (57.9%) resolved upward. A rule that always calls UP, weighted by where player calls actually landed, would have hit 59.7% against the field's 54.3%. The in-game Monkey Index, an equal weight basket of eleven seeded coin flip monkeys used for the bonus pool, stakes every chart at minimum confidence, so its record measures market exposure rather than chart reading. The tougher benchmark is the always-up rule above.

the crowd

Crowd majority provided limited edge over the month. The majority of active players landed on the right side of only 73 of 135 decided charts, while 5 charts split exactly even.

the vanishing edge

Among the 20 players who made at least 20 calls in each fortnight of the month, the correlation between fortnight-one and fortnight-two hit rates was minus 0.03. The top quartile from fortnight one hit 60.6% and then 49.5%. The bottom quartile hit 42.7% and then 49.8%. Both cohorts converged on the coin flip. Twenty players is a small panel, so this can only be an early reading. Even with that constraint, the direction is textbook regression to the mean. Month one's leaderboard measured luck.

coin flip top quartile 60.6% bottom 42.7% 49.5% · 49.8% fortnight one fortnight two
Top-quartile hit rates fell from 60.6% to 49.5% between fortnights, while bottom-quartile hit rates rose from 42.7% to 49.8%.
the week at a glance

The last seven complete days: charts that rose vs fell, with the crowd's hit rate.

Sun 26
57.3%
559 calls
Mon 27
47.0%
701 calls
Tue 28
60.3%
672 calls
Wed 29
44.0%
668 calls
Thu 30
30.9%
598 calls
Fri 31
58.9%
615 calls
Sat 01
54.1%
499 calls

One footnote from the week: on Thursday July 30 all five charts fell, the always-up rule scored 0.0%, and the crowd still hit only 30.9%, because it leaned long anyway.

the chart that fooled everyone
86%
of the 128 players who rated it were on the wrong side of the most fooling chart of the month: PNC Financial, dealt blind from December 2023, which drifted −1.59% over its five days while the crowd leaned the other way. It replaces 3M (81% fooled) from the fortnight report.

The game deals five new blind charts at 00:00 UTC every day.