The Synth coordinator is evaluated on incremental return predictions, not raw prices.
Incremental returns capture the relative change in price and produce a stationary series that is easier to model and compare across assets.

Below is an example of a density forecast over incremental returns for the next 24h at 5-minute intervals:

Below is a minimal example showing what your tracker might return:


  # Expected output

  >>> model.predict(asset="SOL", horizon=86400, step=300)

  [

      {

          "step": (k + 1) * step,

          "prediction": {

              "type": "builtin",

              "name": "norm",

              "params": {

                  "loc": -0.01,       # mean return

                  "scale": 0.4     # standard deviation of return

              }

          }

      }

      for k in range(0, horizon // step)

  ]

Here is the return forecast mapped into price space:

Game Rules

Start

The game begins with a model calibration / warm-up phase of 4 weeks, where predictions are scored but not rewarded.
Leaderboard is based on 7-day rolling average of CRPS scores.
A player must wait 7 days to get a meaningful ranking.
Players may enter or exit the game at any time.
Each player may run two active models, which can be updated at any time.

Prediction Phase

In each prediction round, players must submit a set of density forecasts.

A prediction round is defined by one asset, one forecast horizon and one or more step resolutions.

A 24-hour horizon forecast
- Triggered every 1 hour for each asset
- Step resolutions: {5-minute, 1-hour, 6-hour, 24-hour}
- Supported assets:
  ["BTC", "SOL", "ETH", "XAUT", "SPYX", "NVDAX", "TSLAX", "AAPLX", "GOOGLX"]
A 1-hour horizon forecast
- Triggered every 12 minutes for each asset
- Step resolutions: {1-minute, 5-minute, 15-minute, 30-minute, 1-hour}
- Supported assets:
  ["BTC", "SOL", "ETH", "XAUT"]

All required forecasts for a prediction round must be generated within 40 seconds.

Don't submit models that needs more than 40 seconds to infer!

Scoring

Once the full horizon has passed, each prediction is scored using a CRPS scoring function.
A lower CRPS score reflects more accurate predictions.
Missing or invalid predictions receive the worst CRPS score of the round.

Leaderboard ranking is based on a 7-day rolling average of CRPS scores across all assets and horizons, evaluated relative to other participants, for each prediction round:

The best CRPS score is assigned a normalized score of 1
The worst 5% of CRPS scores are assigned a score of 0
Intermediate scores are scaled accordingly

Leaderboard & Game State

The leaderboard displays several relative performance indicators, computed across all assets and horizons:

Primary Metric
- Anchor CRPS (7 days): Rolling relative CRPS average over the past 7 days.
Secondary Metrics
- Steady CRPS (3 days): Rolling relative CRPS average over the past 3 days.
- Recent CRPS (24 hours): Rolling relative CRPS average over the past 24 hours.

Payouts

$30K over the first 4 months followed by real mining rewards from Synth Miners (currently up to 50K / months)
Rewards are distributed at target resolution + 24h (every 7 + 1 days)
Fixed reward pool is allocated for each payout period.
the top 10 participants receive 100% of the pot.
Only models outperforming the benchmark model (synth:benchmarktracker) are rewarded.
Models with a score below or equal to the benchmark at payout time are included in calculations but receive no payout, leaving any fraction of the pool tied to lower-performing models undistributed.
The benchmark model is a reference model provided by the game in order to reward community outperformance over internal Benchmark and may evolve over time.

Example of reward distribution when 7 people beat the Benchmark.

In this example, only 7 participants outperform the benchmark and receive approximately 70% of the weekly rewards. The remaining 30% is retained in the rewards pool.