Stock chart prediction game: five blind charts every day

A useful stock chart prediction game should test a decision, not recognition. Read the Tape gives every player the same five historical S&P 500 charts, removes the ticker and dates, and normalises the first visible close to 100. You call UP or DOWN for the next five trading days before the answer is revealed.

The outcome is real rather than invented. After each call you see the company, the period, the five-day move and the signals that were visible at the decision point. Confidence sizes a paper-money position, so a lucky guess made with unjustified certainty is still a poor forecast.

What the game actually measures

Directional accuracy is only the first layer. The confidence choice measures whether your certainty matches reality. A 55% call is a lean. A high-confidence call commits most of the paper book. That makes repeated play a test of calibration as well as pattern recognition.

All players receive the same daily set, so crowd splits compare like with like. The Monkey Index provides a deliberately uninformed benchmark made from eleven seeded random callers.

Why the chart is blind

Knowing the ticker or the year invites stories that were unavailable at the decision point. Hiding both removes hindsight clues. The visible candles, moving averages, volume and RSI are computed only from information available at the final visible bar.

The five hidden sessions stay hidden until you commit. That no-lookahead rule is what turns a chart viewer into an honest prediction game.

The subtle ways a chart test leaks the answer

Indicators leak too. A centred moving average, a scaler fit on the whole series, or a chart range set using hidden highs and lows all let future observations influence what the player sees. Signals and display bounds have to stop at the decision bar.

Each daily chart is selected from a deterministic date-and-slot seed and persisted once dealt. Everyone sees the same tape, and a restart cannot quietly replace a difficult chart.

What a clean test does not prove

A no-lookahead design does not show that technical analysis works. It only ensures the question was asked honestly. Selection effects, regime changes and the small sample from five charts a day all still apply, which is why a long calibration record beats a memorable result.

Questions