Coefficients on log terms are elasticities. The ecosystem term is a log-space shift, so its premium is exp(c) − 1. Fitted with a light ridge penalty chosen by leave-one-out (grid 0.1, 0.3, 1, 3, 10), never on the intercept.
Predicted vs actual FDV Listing day Live today
basispoint.xyz
Each point is one observation the selected model was fitted on; the line is y = x. Points above it are venues the model thinks should be worth more than the market paid; below it, less. Both axes are logarithmic — a point one gridline off is wrong by 10×.
Every regressionranked by leave-one-out error
Model
Revenue · ecosystemselected
10
0.65
·
·
1.83×6.3
1.0
0.79
0.73
+145%
×3.3
Volume · ecosystem
11
·
·
0.64
0.90×2.5
10.0
0.50
0.38
+181%
×5.2
Revenue
10
0.67
·
·
·
3.0
0.66
0.62
+196%
×4.1
Where each programme should listregression vs multiples model
—Not available for this venue; the regression that priced it does not use this input.
—Not available for this venue; the regression that priced it does not use this input.
—Not available for this venue; the regression that priced it does not use this input.
no
no inputs
—
—
Observations8 listing-day · 8 live
HyperliquidHYPE
listing day
2024-11-29
$6.00B
—
$2.24B
$2.70B
yes
GRVTGRVT
listing day
2026-07-30
$231.99M
—
$360.00M
$1.20B
no
Open interest · ecosystem
11
·
1.06
·
0.86×2.4
0.1
0.58
0.48
+206%
×4.5
Open interest + Volume · ecosystem
11
·
0.41
0.67
1.20×3.3
3.0
0.65
0.50
+220%
×4.4
Open interest
11
·
1.02
·
·
1.0
0.54
0.49
+231%
×4.5
Volume
11
·
·
1.15
·
0.1
0.49
0.44
+239%
×4.9
Open interest + Volume
11
·
0.42
0.40
·
10.0
0.50
0.37
+256%
×5.3
Revenue + Volume
7
0.37
·
0.64
·
10.0
0.46
0.19
+260%
×6.7
Revenue + Open interest
7
0.39
0.29
·
·
10.0
0.45
0.18
+303%
×6.7
Revenue + Open interest + Volume
7
0.33
0.23
0.52
·
10.0
0.48
-0.03
+410%
×8.5
Why fit fourteen models instead of one: programmes are missing things. A zero-fee venue has no revenue to take a log of; a venue the funding API does not carry has no OI or volume. Rather than impute, every subset is fitted and each programme is priced by the best model its own data can feed. Note how the richer models score worse out of sample — with a dozen observations and three inputs that all measure “how much trading”, extra variables mostly buy overfitting. That is why selection is by LOO error, never by variable count.
—Not available for this venue; the regression that priced it does not use this input.
$606.12M
$674.90M
no
$646.46M$123.67M – $3.38B
Vol · ecosystemn=11, R² 0.50, LOO +181%. 6 fits were applicable.
$739.78M
$474.57M
+36%
“Model” is the regression that priced the venue — the applicable fit with the lowest LOO error. “Ensemble” is the median across every applicable fit, a check on whether the chosen one is an outlier. Revenue for unlisted programmes is the fee schedule applied to volume, which overstates realised take; treat revenue-only estimates for high-fee venues with that in mind.
dYdXDYDX
listing day
2021-09-08
$10.65B
$78.00M
—
—
yes
LighterLIT
listing day
2025-12-30
$2.79B
$113.10M
$1.60B
$4.70B
no
AsterASTER
listing day
2025-09-17
$3.88B
$61.61M
$254.00M
$1.50B
no
EdgeXEDGE
listing day
2026-03-31
$673.38M
$190.61M
$1.10B
$3.00B
no
DDriftDRIFT
listing day
2024-05-16
$300.83M
$11.63M
—
—
no
ParadexDIME
listing day
2026-03-05
$72.19M
$7.57M
$672.00M
$1.17B
no
HyperliquidHYPE
live
2026-10-23
$74.44B
$786.67M
$11.51B
$6.22B
yes
GRVTGRVT
live
2026-10-23
$234.82M
—
$439.07M
$488.79M
no
dYdXDYDX
live
2026-10-23
$111.35M
—
$47.49M
$58.98M
yes
LighterLIT
live
2026-10-23
$3.40B
$34.57M
$557.10M
$2.22B
no
AsterASTER
live
2026-10-23
$5.45B
$50.17M
$1.35B
$1.87B
no
EdgeXEDGE
live
2026-10-23
$325.29M
—
—
—
no
DDriftDRIFT
live
2026-10-23
$14.88M
—
—
—
no
ParadexDIME
live
2026-10-23
$44.02M
$237.35K
—
—
no
Listing-day rows are the comparables set: FDV on TGE day against that day's metrics. Live rows are refreshed daily: today's FDV from CoinGecko against today's OI and volume from the funding backend and trailing-30-day revenue from DefiLlama. A live metric under 5% of the venue's TGE figure is treated as partial upstream coverage and left blank rather than fed to the fit.
AdrenaNo points programme
HyperliquidHYPE
GRVTGRVT
