| Horizon | High (P90) | Median | Low (P10) |
|---|---|---|---|
| 30 days | $3,098 increased by 38.60% from reference close | $2,335 increased by 4.50% from reference close | $1,794 decreased by 19.70% from reference close |
| 90 days | $4,395 increased by 96.60% from reference close | $2,524 increased by 12.90% from reference close | $1,449 decreased by 35.20% from reference close |
| 1 year | View 1-year illustrative outlook | ||
Rounded USD prices. Low and high are model P10 and P90; changes are from each snapshot’s reference close.
Hover or tap to inspect. Use arrow keys to explore, or Escape to dismiss.
The bullish scenario is the model P80 terminal estimate, with its P70-P90 corridor shown as a range.
The bearish scenario is the model P20 terminal estimate, with its P10-P30 corridor shown as a range.
This stressed downside marker extends the fitted loss tail beyond ordinary simulations and should not be read as a precise probability forecast.
Probabilistic scenario analysis, not a guarantee or investment recommendation.
| Year | Target date | Upside scenario | Central scenario | Downside scenario |
|---|---|---|---|---|
| 2026 | Year-end UTC close | $3,106 increased by 39.00% from reference close | $2,257 increased by 1.00% from reference close | $1,630 decreased by 27.10% from reference close |
| 2027 | Year-end UTC close | $5,919 increased by 164.80% from reference close | $2,378 increased by 6.40% from reference close | $955.07 decreased by 57.30% from reference close |
| 2028 | Year-end UTC close | $8,159 increased by 265.10% from reference close | $2,404 increased by 7.60% from reference close | $690.73 decreased by 69.10% from reference close |
| 2029 | Year-end UTC close | $9,908 increased by 343.30% from reference close | $2,373 increased by 6.20% from reference close | $502.15 decreased by 77.50% from reference close |
| 2030 | Year-end UTC close | $12,795 increased by 472.50% from reference close | $2,353 increased by 5.30% from reference close | $423.54 decreased by 81.10% from reference close |
Illustrative USD scenarios for December 31 through 2030. Percentage changes use the frozen reference close. A simulation uses up to three years of price history and drift shrinkage, adjusted by saved milestone assumptions. These bounds are not calibrated probability intervals; short-term backtests do not validate the annual model.
Reference close: Nov 6, 2026. Each event effect phases in once and is then held constant. Neutral central assumptions leave the central baseline unchanged. Overlapping events and effects already reflected in price can distort results. Review the assumptions with each event; price effects are illustrative editorial inputs.
We use historical Ethereum prices to model possible outcomes. Backtested forecasts and fitting-only illustrative outlooks have separate evidence and limitations.
Past price data informs volatility and market patterns.
Models estimate possible future prices. Illustrative outlooks also apply saved milestone assumptions.
Held-out forecasts are scored against a simple baseline. This does not validate the separate illustrative outlooks.
CRPS improvement +0.00%; one-sided 95% lower bound -1.28%.
The forecast combines a horizon-aware, drift-shrunk Student-t GARCH model, filtered historical simulation, and regularized quantile regression. Nonnegative model weights are learned from earlier Monday-UTC forecast dates without using future data. An embargo of 4 weekly origins separates calibration from evaluation, followed by 20 held-out forecasts spaced 35 days apart. Performance is measured using CRPS and coverage diagnostics against the strongest predeclared simple baseline. An EVT model extends the downside tail for the black-swan stress scenario.
Price data CoinMarketCap, CryptoSlate
Momentum shifts the regularized quantile component while shrinkage limits extrapolation.
Higher conditional volatility widens both upside and downside predictive bands.
The first release retains a zero-weight boundary until historical overlay validation is available.
Past market behavior may not represent a new market structure or regulatory shock.
The forecast targets one UTC terminal price and does not predict the full intraday path.
Actual prices can finish outside every displayed interval.
Non-overlapping labels reduce mechanical overlap, but regime dependence can persist and manual review remains required.
An origin is a historical date when the model is tested as if it were issuing a live forecast.
Public raw backtest origins are fixed to Monday UTC to prevent schedule shopping, meaning results cannot be improved by choosing unusually favorable start dates. This horizon requires at least 1,534 continuous daily candles, including up to six days of weekday-alignment slack.
Calibration origins are earlier forecasts used to choose the ensemble weights. Held-out origins are kept separate and used only to evaluate the chosen model.
This run used 72 weekly calibration origins and compared weight selection with Student-t GARCH. An embargo then skipped 4 weekly dates, creating a deliberate gap before evaluation.
The final 20 held-out origins were issued 35 days apart across a 96-position validation span so their 30-day outcomes do not overlap.
CRPS stands for Continuous Ranked Probability Score. It measures the accuracy of the full probability forecast rather than a single price target; lower is better because it rewards both an accurate center and honest uncertainty bands.
The held-out point edge against the strongest simple baseline was 0.0%. The minimum one-sided 95% paired lower bound was -1.3%; this conservative check asks whether the measured advantage remains above zero after accounting for sample variation.
The monitoring target is a point edge of at least 2% and a positive lower bound against both GARCH and filtered historical simulation. These regime-dependent diagnostics remain advisory and do not interrupt publication.
A simulation path is one plausible 30-day price journey drawn by the model.
The published run used 2,000 paths for the displayed distribution and 2,000 paths at each backtest origin.
RNG seed 1,047,148,897,491,693,997 fixes the random-number stream so an identical run can be reproduced. It is deterministically derived from the frozen forecast.v1 production salt, horizon, asset ID, and modeled reference-close timestamp; candles, provenance, and market inputs do not select the random stream. The modeled reference close is timestamped 2026-11-06T23:59:59Z.
No quality-qualified Polymarket evidence contributed to this forecast, so prediction-market prices did not move the published distribution.
Showing retained 30-day forecasts. 2 of 2 checked closes fell inside the 80% model range (100%).
Scroll to view all columns
| Forecast date | Target date | Median | 80% range | Actual close | Result |
|---|---|---|---|---|---|
| $2,495.48 | $1,920.59 to $3,195.20 | $2,705.54 | Inside range | ||
| $2,586.25 | $1,810.50 to $3,618.74 | $2,752.63 | Inside range |
One first-retained forecast per two-month calendar period, up to six rows. Completed checkpoints take priority. Values stay fixed; results use exact CoinMarketCap USD daily closes. The displayed sample’s hit rate excludes pending targets and missing closes. Past results do not guarantee future accuracy.
What public figures said, when they said it, and the targets they set. These statements are separate from the CryptoSlate forecast.
Hayes maintained a $10,000 ETH target for year-end 2026 in his September 2 outlook, conditional on financial stress driving additional dollar liquidity. CryptoSlate later reported that he closed Maelstrom’s directional crypto positions on October 24. That change in positioning does not itself establish that he formally withdrew the price target.
CryptoSlate reported on October 19, 2025 that Tom Lee had reaffirmed a $12,000 to $15,000 ETH range by year-end 2025, citing tokenization, DeFi, AI infrastructure and improving liquidity. The precise date of the underlying statement was not established in that coverage.
ETH did not reach the $12,000–$15,000 range by December 31, 2025. CoinMarketCap records ETH’s 2025 high at $4,953.73 on August 24, below the range’s $12,000 lower bound. This assessment applies to the historical price range and year-end deadline, with tokenization, DeFi, AI infrastructure and improving liquidity cited as the call’s supporting rationale. Assessment evidence · Reviewed Nov 7, 2026
Kendrick projected ETH at $7,500 by the end of 2025, arguing that corporate-treasury buying and ETF inflows could tighten available supply. This records the historical 2025 call, rather than a current price target.
The $7,500 target was not reached by December 31, 2025. CoinMarketCap records ETH’s 2025 high at $4,953.73 on August 24, below the target throughout the year. This assessment applies to the historical price target and year-end deadline; corporate-treasury buying and ETF inflows were the supporting rationale for that call. Assessment evidence · Reviewed Nov 7, 2026
VanEck’s research team, led by Matthew Sigel, projected ETH at roughly $22,000 by 2030 if Ethereum retained its leading position among smart-contract platforms. The model assumed about $66 billion in annual free cash flow and a 33-times valuation multiple.
Historical statements retain their original context. Targets may change; inclusion is not endorsement or a claim of accuracy.
Backtested forecasts model prices at a stated target date using historical returns and volatility. Separately labeled illustrative outlooks fit scenarios to available price history with any saved milestone assumptions. The Method section identifies the inputs and validation evidence for the selected run.
In the backtested forecast table, low and high are the 10th and 90th model percentiles at the target date. Actual outcomes can fall outside them. Separately labeled illustrative outlooks use downside, central and upside scenarios with editorial milestone assumptions; their bounds are not calibrated probability intervals.
Enough daily price history is available to fit a one-year model, but held-out one-year validation is unavailable. Its upside, central and downside scenarios are illustrative, not calibrated probabilities. A shorter-horizon backtest does not validate this outlook.
The one-year target is 365 days after its reference close. By-year rows are December 31 UTC closes, so the target dates can differ. The illustrative one-year and year-end targets use the same frozen simulation and saved milestone assumptions.
Check the Method section and Track record where available. Backtests compare earlier forecasts with known outcomes and a simple baseline; a result for one horizon does not validate another. Historical performance does not establish future accuracy, and annual scenarios and illustrative outlooks have separate limitations.
No. A terminal distribution describes prices at a target date. It does not give the probability of touching a threshold at any point before that date.
Each forecast uses a frozen reference close. Current quotes can change after the run. Forecast percentage changes are measured from that reference close, not from the latest quote.
Price targets require enough usable historical data and an available published model output. A horizon can also be paused or withheld. Sections such as annual forecasts, milestones and prediction markets are omitted when they have no published content.
No. Model assumptions, limited historical samples and changing market conditions create substantial uncertainty. Older snapshots are labeled, and paused forecasts withhold current targets.
