argus

ETH/USD Log Returns 8h

as of 22:07 UTC · refreshes hourly
Latest answer
+0.099%
block 11,987,797
Paid out here · per day
≈632.6 ALLO/day
to the 17 models, 6 forecasters and 5 reputers working here, split by score
Reward share
6.92%
Models · 7d
17
Topic stake
520,753
ALLO staked by reputers here
Epochs shown
200
129 with truth
How this topic scores
Loss method
czar
Truth arrives
11.0 h later
Price source
not declared on chain

CZAR, composite zero-agnostic return: an asymmetric piecewise loss built to stop constant near-zero calls from scoring well, which plain error rewards. The topic definition names an asset, a kind and a horizon and nothing else, so each reputer sources the truth privately and the settled answer is their stake-weighted agreement. Our own charts score against spot candles, which is an independent measurement rather than the chain's.

Network answer vs what actually happened combined naive mean actual
actual is our independent reconstruction from spot candles: exact for log returns and prices, estimator-dependent for volatility · the shaded 8h has no actual yet, a call is only scorable once its horizon has passed
Is the network getting better here combined naive
rolling 20-epoch mean error vs our reconstruction · lower is better · the gap between the lines is the forecaster layer's edge
Forecaster influence
rolling 20-epoch mean of how far the synthesis moved off the plain mean
How money moves through this topicone way in, one way out
1 · Money in

Funding. It is spent, never returned, and buys the topic a bigger share of network emissions.

2 · The pool

That share is 6.92% of emissions, which is what the daily figure below measures.

3 · Money out

Paid to the models, forecasters and reputers working here. Backing a reputer is how you take a share without running a model.

Paid out here, per day
≈632.6 ALLO/day

Set by fee revenue and reputer stake, and split by score among everyone working the topic. Two ways to be involved, and they are opposites: funding pays in and buys you a better answer, backing a reputer earns a share of what goes out.

Funding history
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-20
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-18
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-17
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-16
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-15
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-14
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-13
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-12
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-11
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-10
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-09
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-08
allo1jxtw93k38t5tk…6v60100 ALLO2026-11-07
allo1ft03vkprf5upv…6geg100 ALLO2026-11-06
allo1ft03vkprf5upv…6geg100 ALLO2026-11-05
Forecaster edge · does the synthesis beat the plain mean here
3.1% bettercombined MAE 0.0339 vs naive 0.0350 · combined closer on 67% of 129 scored epochs
Model seats
17of 32
15 paid slots open
Forecaster seats
6of 6
full, a new entrant earns nothing here
Paid to ask
652
ALLO left in this topic's pot right now, read from the chain. It drains as rounds pay out, so it is smaller than the sum of the payments below.
Worker leaderboard · last 7 daysour error (raw) beside the chain's own score
#WorkerEpochsAttendanceMean errorChain score
1allo1rzlpcck2f…2a7e1,462100.0%0.01623#16-0.0225
2allo1emx6phjc0…ek861,44398.7%0.01627#9-0.0092
3allo1mmx6a94xr…hgep1,462100.0%0.01645#10-0.0096
4allo1s90m0g464…cyje1,44799.0%0.01672#2+0.0430
5allo1h78vpjv4e…gr7l1,44098.5%0.01716#1+0.1269
6allo1vjenmm4y9…px3c1,46199.9%0.01756#15-0.0197
7allo1d2yt7846a…6clu1,43097.8%0.01853#7-0.00063
8allo1jlyxfymwx…5tt31,462100.0%0.01885#6+0.0033
9allo15jznxdhrx…9a3w1,462100.0%0.01896#5+0.0042
10allo1ck89mk0f5…pl5x1,43898.4%0.01915#13-0.0128
11allo17lk6v4ad0…tvua1,43998.4%0.01916#4+0.0097
12allo1s2gt55kfg…ct7n1,462100.0%0.01983#12-0.0115
13allo18u3dygwvq…78g049433.8%0.02092#17-0.1558
14allo1y8xy5rpvl…87qd1,462100.0%0.02194#3+0.0107
15allo105fuhxgxe…gztu1,462100.0%0.02248#8-0.0052
16allo1h4grarj6r…lcv81,462100.0%0.02291#11-0.0101
17allo1t6fns6ej3…70r31,462100.0%0.02313#14-0.0134

17 of 17 workers scored. Mean error is against our independent reconstruction from spot candles: it asks whether a prediction was close to what happened. Ranks on volatility topics inherit that estimator's bias and are directional only. Chain score is Allora's own, read from the emissions module: it asks how the network rated a worker against the others, and it is the number that decides pay. The two disagree, and that is the point of showing both.