Executive Function
Streak Prediction Task
Also called: gambler's fallacy task, gambler's fallacy and hot hand task, negative recency task, binary sequence prediction task, next-outcome prediction task
Measure how predictions of a chance event change as a streak grows: participants predict each next color of a light that is fair every time, scored as the streak slope.
5-8 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Try it here first, with no account.
- Papers, 2021 to 2025
- 208
- Evidence grade
- C
- Registered clinical studies
- 0
What it measures
Measures how predictions of a chance outcome change as a streak grows. A light shows orange or blue, each equally likely every time, and participants predict each next color. The main score is the streak slope: the change in the share of predictions that a streak will end, for each color added to it.
Constructs
- Judgment under uncertainty
- Beliefs about randomness
- Sequence prediction
- Negative recency
Research uses
- Research on judgment under uncertainty and beliefs about chance
- Decision-making and cognitive bias research
- Within-study comparison of predictions after streaks
How the paradigm works
The streak prediction task measures how predictions of a chance event change as a run of the same outcome grows. The gambler's fallacy is the belief that, for random events, a run of one outcome will be balanced by the other (Ayton and Fischer, 2004), and Tversky and Kahneman (1971) described the wider intuition that even a small random sample should look like the population it comes from. Jarvik (1951) found that predictions of the next of two symbols turned against a run once it passed two, more so as it grew (negative recency). Ayton and Fischer (2004) found that the longer the run of a color on a simulated wheel, the less often people predicted that color, a linear trend by run length, and Xue and colleagues (2012) measured the pattern with guesses of a random red or black card after streaks of 4 or more.
More about the paradigm
A light drawn by code shows orange or blue. Each color is an independent draw with probability one half from the session's seed, saved on every row, and the instructions say that both colors are always equally likely, whatever came before. The latest 8 colors show in a row above the light. In each round the participant predicts the next color by click, tap, or the F and J keys, with no time limit; the two buttons take their sides from a seeded draw, so a change of color is not a change of hand, as in Xue and colleagues' task. The light then shows the color, with a line saying whether the prediction was right. There are 3 practice predictions and 100 rounds (60 to 200 in the settings). About 5 to 8 minutes.
The main score is the streak slope: the change in the share of predictions that the streak will end (reversals), for each color added to the streak, from 1 to 5 or more. A positive slope means more reversal predictions as streaks grow, the gambler's fallacy pattern; a negative slope means more predictions that streaks go on. The share of reversals after streaks of 4 or more, Xue and colleagues' measure, comes next, then the shares after streaks of 1, 2 and 3. With a fair light about 1 round in 8 follows a streak of 4 or more, so one person's long-streak share rests on few rounds. No way of predicting is better than another: every way is right about half the time. One color, or one side, on at least 90% of rounds is flagged for review. Scores support research and do not diagnose.
Scores
- Streak Slopescore
- Least-squares slope of reversal predictions (1) against continuations (0) on the streak length, 1 to 5 with longer streaks pooled: the change in the share of reversal predictions for each color added to the streak.
- Positive: more reversals as streaks grow; negative: more continuations; 0 is no change
- Reversals After Long Streaks%
- Share of predictions of the other color after streaks of 4 or more.
- 0.5 is no lean; above 0.5 expects streaks to end
- Reversals After One%
- Share of predictions of the other color after a streak of 1.
- Informational
- Reversals After Two%
- Share of predictions of the other color after a streak of 2.
- Informational
- Reversals After Three%
- Share of predictions of the other color after a streak of 3.
- Informational
- Reversal Rate%
- Share of predictions of the other color than the last, over every round.
- Informational
- Most Predicted Color%
- Share of rounds on the color predicted more often.
- 0.9 or more is flagged for review
- Most Chosen Side%
- Share of rounds on the button side chosen more often; the colors change sides at random.
- 0.9 or more is flagged for review
- Median Decision Timems
- Median time to a prediction, at or after 100 ms.
- Informational
- Anticipationscount
- Presses within 100 ms of a round appearing. They are not taken as predictions.
- Quality check
- Rounds Predictedcount
- Predictions in the scored run.
- Fewer than planned is flagged for review
- Planned Roundscount
- Rounds set for the session.
- Informational
Use and evidence
208 papers from 2021 to 2025 name this task in their title or abstract, up 9% compared with 2016 to 2020. That ranks it 138 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 58 |
| 2024 | 47 |
| 2023 | 30 |
| 2022 | 34 |
| 2021 | 39 |
| 2020 | 18 |
| 2019 | 53 |
| 2018 | 52 |
| 2017 | 35 |
| 2016 | 32 |
| 2015 | 39 |
| 2014 | 37 |
| 2013 | 32 |
| 2012 | 20 |
| 2011 | 27 |
| 2010 | 19 |
| 2009 | 21 |
| 2008 | 20 |
| 2007 | 9 |
| 2006 | 13 |
| 2005 | 9 |
| 2004 | 10 |
| 2003 | 11 |
| 2002 | 9 |
| 2001 | 4 |
| 2000 | 1 |
Use grade C
100 to 299 papers from 2021 to 2025.
Evidence grade C
At least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies.
- Papers on test-retest reliability
- 1
- Papers on reliability or validity
- 34
- Registered clinical studies using it as an outcome
- 0
Note on the counts: includes betting-market papers that run no task.
How the grades work
Use: A is 1,000 or more papers from 2021 to 2025, B 300 to 999, C 100 to 299, D 30 to 99 and E fewer than 30. Evidence: A means at least 25 papers report its test-retest reliability and at least 25 registered clinical studies use it as an outcome; B, at least 8 test-retest papers, and either 5 registered clinical studies or 150 reliability or validity papers; C, at least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies; D, little published measurement evidence found. The counts measure use. They include every paper that names the task in its title or abstract. Grades describe the published research on this task, not ConductCognition's version of it.
Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Light
- A round light drawn by code shows orange or blue. Each color is an independent draw with probability one half from the session's seed, and the instructions say that both colors are always equally likely, whatever came before. The whole sequence is saved on every row.
- Row
- The latest 8 colors show above the light, newest on the right; 8 colors are shown before the first prediction.
- Each round
- The participant predicts the next color by click, tap, or the F and J keys, with no time limit. The two buttons take their sides from a seeded draw in each round. The light then shows the color, with a line saying whether the prediction was right, for 1,000 ms, then a 400 ms blank. Presses under 100 ms are not taken.
- Rounds
- 100 (standard; 60 to 200), after 3 practice predictions of the same light. About 1 round in 8 follows a streak of 4 or more.
- Scores
- The streak slope (the change in the share of predictions that a streak will end, for each color added to it, with streaks of 5 or more pooled), the share after streaks of 4 or more, and the shares after streaks of 1, 2 and 3.
Xue and colleagues (2012) used red and black cards, money for each right guess and a sequence set in advance. This test uses a light and colors of its own, no money, and a new fair sequence for each session, so its rates are not directly comparable with published ones. Streaks of 4 or more are few in one session, so compare groups. No pattern of predictions is better than another: every color is equally likely. The scores support research and do not diagnose.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: streak_prediction_round from the scored run, one prediction of orange or blue per round with no time limit. Every color is a fair draw, so no prediction is right more often than half the time and accuracy is not judged against chance. One color on at least 90% of rounds, one side on at least 90% of rounds and a run short of its planned rounds are flagged for review. Predictions under 200 ms on at least half the rounds invalidate.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. The light is fair, so a prediction is a belief about the next color and is never wrong in advance. Predicting one color nearly every round, or pressing the button on one side while the colors change sides at random, means the predictions did not depend on the streaks; both are choices, so they are flagged and never exclude. An answer under 200 ms is faster than a simple visual reaction, so it cannot reflect the item. Following response-time effort screening for untimed tests (Wise and Kong 2005), the session is invalid only when at least half the answers are that fast.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Rounds | 100 | 60 to 200 |
Yearly plans can change settings within these limits. Free and pay-as-you-go studies use the standard settings.
Need a setting this test does not have?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Languages: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese.
- Available on: Every plan, including the free plan.
- No percentile. Scores and quality flags only. How norms work
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Streak Prediction Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Jarvik, 1951; Tversky & Kahneman, 1971; Ayton & Fischer, 2004), with the standard settings (rounds: 100). The main score was streak slope (least-squares slope of reversal predictions (1) against continuations (0) on the streak length, 1 to 5 with longer streaks pooled: the change in the share of reversal predictions for each color added to the streak).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Jarvik, M. E. (1951). Probability learning and a negative recency effect in the serial anticipation of alternative symbols. Journal of Experimental Psychology, 41(4), 291-297. https://doi.org/10.1037/h0056878Predictions of the next of two symbols that turned against a run once it passed two, more so as it grew (negative recency).
- Tversky, A., & Kahneman, D. (1971). Belief in the law of small numbers. Psychological Bulletin, 76(2), 105-110. https://doi.org/10.1037/h0031322Erroneous intuitions about chance: a short random sample expected to look like the population it comes from.
- Ayton, P., & Fischer, I. (2004). The hot hand fallacy and the gambler's fallacy: Two faces of subjective randomness? Memory & Cognition, 32(8), 1369-1378. https://doi.org/10.3758/BF03206327Predictions of a simulated wheel's red and blue outcomes: the longer the run of a color, the less often it was predicted, a linear trend by run length.
- Xue, G., He, Q., Lei, X., Chen, C., Liu, Y., Chen, C., Lu, Z.-L., Dong, Q., & Bechara, A. (2012). The gambler's fallacy is associated with weak affective decision making but strong cognitive ability. PLoS ONE, 7(10), e47019. https://doi.org/10.1371/journal.pone.0047019The design: guesses of a random red or black card with the last five shown and the answers' sides varied, scored after streaks of 4 or more.
- Studer, B., Limbrick-Oldfield, E. H., & Clark, L. (2015). 'Put your money where your mouth is!': Effects of streaks on confidence and betting in a binary choice task. Journal of Behavioral Decision Making, 28(3), 239-249. https://doi.org/10.1002/bdm.1844Red and black predictions that turned against color runs in three laboratory studies.
Source papers
- Xue G, He Q, Lei X, Chen C, Liu Y, Chen C, Lu Z-L, Dong Q, Bechara A (2012). The gambler's fallacy is associated with weak affective decision making but strong cognitive ability. PLoS ONE, 7(10):e47019.N 438, College students
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Streak Prediction Task take?
About 5 to 8 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops, tablets and phones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Streak Slope, Reversals After Long Streaks, Reversals After One, Reversals After Two, Reversals After Three, Reversal Rate, Most Predicted Color, Most Chosen Side, Median Decision Time, Anticipations, Rounds Predicted and Planned Rounds. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Streak Prediction Task used in research?
208 papers from 2021 to 2025 name it in their title or abstract, up 9% compared with 2016 to 2020. That ranks it 138 of 495 test families in the ConductCognition test census (use grade C). Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Streak Prediction Task have?
In the published research, 1 paper reports its test-retest reliability and 0 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Rounds (60 to 200). Settings can be changed on yearly plans; free and pay-as-you-go studies use the standard settings.
Which languages can participants use?
Participants can take this test in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese. English is on every plan, and the other languages come with yearly plans. Have a native speaker read the machine-translated text before launch.
Is this a clinical test?
No. ConductCognition is for research use only. It runs, scores and exports the test; clinical interpretation is outside the platform.
Every test: the full test list.
Run this test in your next study
Start free with 10 free participants on any test, then use participant credits or a yearly plan.