Neuron registration trends
The alpha price in τ, day by day, and what moved it each way.
Log change over the last 30 days −3.65%: protocol buys 0.00%, root sells −0.86%, users net −3.23%, cross-subnet +0.44%, unattributed −0.0002%.
The chart follows the alpha price in τ. Each period starts at its opening price and steps through the mechanisms below, in a fixed order, to its closing price: a step that rises pushed the price up, one that falls pushed it down, and its height is how far it moved the price. A thin line carries each close into the next period’s open; it is dashed where the two differ. The faint candle behind each period is the market’s own open, high, low and close, for context.
With many periods on screen (90 days, or hours), the steps would be too thin to read, so the chart draws the price as a line instead and colours each period by the mechanism that moved it most. Hover a period for the full step-by-step breakdown.
Not here, because they do not move the spot price: the per-block TAO injection into the pool, and burns.
All figures are log changes, not simple percentage changes, so the parts add up exactly to the total; on big moves the two differ (a 10% log fall is a 9.5% price fall).
2026-11-11 UTC, as of ≈15:17 UTC (block 9,483,874)
The day isn’t over yet, so these figures keep changing. They are in no daily total, average or rank.
Holder & circulating-alpha metrics as of (daily close)
Financial institutions and hedge funds rely on centralized prediction aggregation systems to rank signal providers. These systems suffer from information leakage before evaluation, concentrated scoring authority, and high operational overhead to onboard diverse prediction sources across multiple timeframes and asset classes. Decentralized scoring networks address this by using cryptographic timelocks to prevent early observation of predictions, distributing validation across independent operators, and creating transparent scoring rules that reward prediction accuracy directly. MANTIS (Bittensor Subnet 123) implements this as a distributed price prediction oracle: miners submit encrypted embedding vectors and scalar predictions across 12 challenges (binary direction, volatility regimes, cross-sectional ranking, range breakouts, and capital-at-risk bracket trades). Validators independently decrypt submissions after a one-week timelock maturity, pair embeddings with on-chain prices, and compute per-miner salience scores using walk-forward logistic regression, permutation importance, and challenge-specific loss functions.
Final weights aggregate across challenges with EMA smoothing and set on-chain every 360 blocks, creating a transparent, auditable signal aggregation system.
Auto-generated from the GitHub repository and may not reflect the latest changes.
MANTIS is Bittensor Subnet 123, a signal aggregation subnet that scores miners on their ability to predict price movements and market dynamics across multiple asset classes and timeframes. Validators collect encrypted price predictions from miners, decrypt them after a timelock maturation window, and compute salience scores using walk-forward machine learning models (logistic regression, LBFGS classifiers, permutation importance) across 12 active challenges including binary direction prediction, volatility regime forecasting, cross-sectional ranking, range breakout detection, and capital-at-risk bracket trading. Final weights are computed via EMA-smoothed challenge aggregation and set on-chain every 360 blocks.
Miners encrypt payload vectors via X25519 ECDH + ChaCha20-Poly1305 and Drand IBE timelock; payloads are committed to Cloudflare R2 and challenge round links downloaded by validators. Validators sample miner payloads every 5 blocks, store encrypted ciphertexts in SQLite ledger, and after PAYLOAD_MATURITY_BLOCKS (50,400 ≈ 1 week) decrypt via Drand public beacon signatures. Decrypted embeddings and on-chain price data are paired in ledger, stratified into walk-forward training/validation folds.
Periodically (WEIGHT_CALC_INTERVAL=1000 blocks), multi_salience() dispatches to per-challenge loss_func (binary, lbfgs, hitfirst, range_breakout_multi, xsec_rank, funding_xsec, flow), computing per-miner salience vectors. Per-challenge weights are aggregated via challenge weighting and EMA-smoothed (alpha=0.3), then set on-chain via subtensor every WEIGHT_SET_INTERVAL=360 blocks. For FLOW bracket trades, miners post bets on the FlowCollateralPool smart contract (Bittensor EVM, chain 964, address 0xD9c805202b16671A2901307fBC9A8750E2453427); validators resolve trades against multi-venue klines, apply path-penalty and Kelly weighting, gate payment behind a rolling significance t-stat, and compute weekly zero-sum settlement (team runs owner settlement daemon separately, not in repo).
Per-challenge scoring paths: (1) Binary: ElasticNet meta-model on walk-forward OOS base-model predictions, feature selection by AUC, importance = |β|. (2) LBFGS: Consensus-merged signals (corr > 0.85) scored via classifier path (permutation importance of balanced accuracy) + Q-path (12 binary models per sigma threshold), blended 75/25. (3) HITFIRST: Two L2 logistic regressions (up/down barrier hit), importance = |β_up| + |β_down|. (4) MULTI-BREAKOUT: Per-event AUC gate (≥0.5, ≥2 episodes), L2 logistic on z-scored episode-balanced predictions. (5) XSEC-RANK: Cross-sectional binary (return > median), walk-forward with recency weighting, importance = |β| * max((AUC_meta - 0.5)/0.5, 0). (6) FUNDING-XSEC: Same as XSEC-RANK on funding rate changes; stale miners zeroed. (7) FLOW: Per-trade path-penalized R, Kelly-weighted into per-regime EWMA, payment gated by significance (max of full-history and rolling last-4-block t-stat on 168h sums, gate z_in=1.25, z_out=0.5); zero-sum weekly settlement, 2% dust tier, remainder burns. Final weight = (1/Σw_c) * Σ w_c * ŝ_{j,c} with EMA alpha=0.3.
The sections above were auto-generated from the GitHub repository and may not reflect the latest changes. View repository (github.com, opens in a new tab, external site)
Incentivizing cooperative prediction
https//mantis123.com/dashboard/
This information was auto-extracted from the GitHub repository and may be incomplete or out of date. Always refer to the subnet's GitHub repository (github.com, opens in a new tab, external site) for the latest instructions.
Not found in the repository docs.
Not yet verified: GPU, Notes
Machine-extracted from the subnet's repository docs; may be wrong or out of date. Check the repository before buying or renting hardware. Open the repository (github.com, opens in a new tab, external site) · Extracted 9 Nov 2026
Install core dependencies including bittensor, torch, scikit-learn, and encryption/timelock libraries
pip install -r requirements.txt
Build timelock 0.0.2.dev0 and wasm wrapper from GitHub; do not pip install timelock (PyPI version is incompatible with current subnet payloads)
./install_reqs.sh
Review challenge-specific payload format and submission requirements
See MINER_GUIDE.md for submission details per challenge type
Use supported CLIs to generate, encrypt payloads, and post bets; no bundled interactive UI(optional)
python generate_and_encrypt.py && python flow_post.py
Install dependencies: bittensor, torch, scikit-learn, scipy, numpy, web3, boto3, and data handling libraries
pip install -r requirements.txt
Build timelock libraries matching validator decryption requirements
./install_reqs.sh
Set environment variables for wallet, S3 bucket credentials, and Drand API access
Configure .env with Bittensor wallet, RPC endpoints, and optional MANTIS_ALLOW_CUDA=1 to enable GPU
Start validator using pm2 process manager
pm2 start validator
Releases distributed as zip archives, not via git; manual update process preserves .env and datalog files
To update: pm2 stop validator, unpack new archive over old directory, run ./install_reqs.sh if needed, pm2 restart validator
| Day (UTC) | Share |
|---|---|
| 2026-11-10 | 0.0003% |
| 2026-11-09 | 0.0003% |
| 2026-11-08 | 0.0003% |
| 2026-11-07 | 0.0004% |
| 2026-11-06 | 0.0006% |
| 2026-11-05 | 0.0006% |
| 2026-11-04 | 0.0006% |
| 2026-11-03 | 0.0006% |
| 2026-11-02 | 0.0006% |
| 2026-11-01 | 0.0006% |
| 2026-10-31 | 0.0006% |
| 2026-10-30 | 0.0005% |
| 2026-10-29 | 0.0005% |
| 2026-10-28 | 0.0005% |
| 2026-10-27 | 0.0005% |
| 2026-10-26 | 0.0005% |
| 2026-10-25 | 0.0005% |
| 2026-10-24 | 0.0005% |
| 2026-10-23 | 0.0006% |
| 2026-10-22 | 0.0006% |
| 2026-10-21 | 0.0006% |
| 2026-10-20 | 0.0006% |
| 2026-10-19 | 0.0006% |
| 2026-10-18 | 0.0006% |
| 2026-10-17 | 0.0006% |
| 2026-10-16 | 0.0006% |
| 2026-10-15 | 0.0006% |
| 2026-10-14 | 0.0007% |
| 2026-10-13 | 0.0006% |
| 2026-10-12 | 0.0006% |
| 2026-10-11 | 0.0006% |
| 2026-10-10 | 0.0006% |
| 2026-10-09 | 0.0006% |
| 2026-10-08 | 0.0006% |
| 2026-10-07 | 0.0006% |
| 2026-10-06 | 0.0006% |
| 2026-10-05 | 0.0006% |
| 2026-10-04 | 0.0005% |
| 2026-10-03 | 0.0006% |
| 2026-10-02 | 0.0006% |
| 2026-10-01 | 0.0005% |
| 2026-09-30 | 0.0006% |
| 2026-09-29 | 0.0006% |
| 2026-09-28 | 0.0005% |
| 2026-09-27 | 0.0005% |
| 2026-09-26 | 0.0008% |
| 2026-09-25 | 0.0007% |
| 2026-09-24 | 0.0006% |
| 2026-09-23 | 0.0006% |
| 2026-09-22 | 0.0007% |
| 2026-09-21 | 0.0007% |
| 2026-09-20 | 0.0007% |
| 2026-09-19 | 0.0007% |
| 2026-09-18 | 0.001% |
| 2026-09-17 | 0.0013% |
| 2026-09-16 | 0.0015% |
| 2026-09-15 | 0.0013% |
| 2026-09-14 | 0.0015% |
| 2026-09-13 | 0.0016% |
| 2026-09-12 | 0.0016% |
| 2026-09-11 | 0.0016% |
| 2026-09-10 | 0.0018% |
| 2026-09-09 | 0.0022% |
| 2026-09-08 | 0.0022% |
| 2026-09-07 | 0.0021% |
| 2026-09-06 | 0.0023% |
| 2026-09-05 | 0.0027% |
| 2026-09-04 | 0.0025% |
| 2026-09-03 | 0.0027% |
| 2026-09-02 | 0.0024% |
| 2026-09-01 | 0.002% |
| 2026-08-31 | 0.002% |
| 2026-08-30 | 0.0018% |
| 2026-08-29 | 0.0019% |
| 2026-08-28 | 0.0019% |
| 2026-08-27 | 0.002% |
| 2026-08-26 | 0.0019% |
| 2026-08-25 | 0.002% |
| 2026-08-24 | 0.002% |
| 2026-08-23 | 0.002% |
| 2026-08-22 | 0.0027% |
| 2026-08-21 | 0.0033% |
| 2026-08-20 | 0.0028% |
| 2026-08-19 | 0.0027% |
| 2026-08-18 | 0.0023% |
| 2026-08-17 | 0.0017% |
| 2026-08-16 | 0.0017% |
| 2026-08-15 | 0.0019% |
| 2026-08-14 | 0.0019% |
| 2026-08-13 | 0.0018% |
