Methodology
How we collected, classified, and analyzed the data.
Data Source
Token listing and delisting history comes from Binance announcements, covering every token whose listing was announced on the blog. Each token is then enriched with current price data, market cap, ATH/ATL, and categories from the CoinGecko API.
Data Filtering
To avoid contaminating metrics, we exclude:
- Stablecoins (USDC, TUSD, FDUSD, etc.) — their price is pegged and ROI is meaningless
- Wrapped/staked tokens (WBTC, stETH, etc.) — they track another asset, not independent performance
Launch Date Determination
Each token's launch date is determined by comparing the Binance listing announcement date with CoinGecko's genesis_date. If the genesis date is earlier than the Binance listing, we use it as the launch date (e.g. DOGE launched in 2013 but was listed on Binance in 2019). Otherwise, the Binance listing date is used.
Dead Token Imputation
Tokens that were delisted from Binance and have no current price data on CoinGecko are imputed with a −100% ROI (total loss). Without this step, dead tokens would be silently excluded from statistical tests, biasing results upward. We do not impute annualized ROI for these tokens since their exact lifespan may be uncertain.
Market Cycle Definitions
Market cycles are defined based on widely-accepted historical BTC price peaks and troughs, aligned with halving cycles. These boundaries are approximate — reasonable people can disagree on exact dates. We include a sensitivity analysis that tests whether shifting boundaries by 1-2 months changes the conclusion.
| Cycle | Start | End | Type |
|---|---|---|---|
| Pre-2013 Early | — | 2013-01-01 | Early |
| 2013 Bull (Spring) | 2013-01-01 | 2013-05-01 | Bull |
| 2013 Correction | 2013-05-01 | 2013-10-01 | Neutral |
| 2013 Bull (Fall) | 2013-10-01 | 2014-01-01 | Bull |
| 2014-2015 Bear | 2014-01-01 | 2015-08-01 | Bear |
| 2015-2016 Recovery | 2015-08-01 | 2016-01-01 | Neutral |
| 2016-2017 Bull | 2016-01-01 | 2018-01-01 | Bull |
| 2018-2019 Bear | 2018-01-01 | 2018-12-01 | Bear |
| 2019-2020 Recovery | 2018-12-01 | 2020-03-01 | Neutral |
| 2020-2021 Bull | 2020-03-01 | 2021-11-01 | Bull |
| 2022 Bear | 2021-11-01 | 2022-11-01 | Bear |
| 2023 Recovery | 2022-11-01 | 2023-10-01 | Neutral |
| 2023-2025 Bull | 2023-10-01 | 2025-11-01 | Bull |
| 2025-2026 Bear | 2025-11-01 | Present | Bear |
Metrics
(current_price - launch_price) / launch_price(1 + token_ROI) / (1 + BTC_ROI) - 1. Positive means it outperformed Bitcoin over the same period.(ATH - current) / ATH. Shows how far the token has currently fallen from its all-time high. Clamped to 0 if current price exceeds ATH.Statistical Tests
We use the Mann-Whitney U test to compare annualized ROI and BTC-relative ROI distributions between Bull and Bear launched tokens. This is a non-parametric test that doesn't assume normally distributed data — appropriate for the heavy-tailed distributions typical of crypto returns.
We report rank-biserial correlation as an effect size measure and compute bootstrap 95% confidence intervals (10,000 resamples) for the difference in medians. A minimum of 20 tokens per group is required for statistical tests.
Important Caveats
- Incomplete delisting data: While we track Binance delistings, tokens that failed on other exchanges or never gained traction are not captured. Some survivorship bias remains.
- Age disparity: Tokens launched in earlier cycles have had more time to appreciate. Annualized ROI and BTC-relative metrics help normalize for this, but imperfectly.
- Sample sizes: Some cycles have fewer tokens. We require a minimum of 20 tokens per group for statistical tests and report sample sizes alongside all results.
- Cycle boundaries are approximate: Different analysts define bull/bear cycles differently. Our sensitivity analysis shows how results change under shifted boundaries.
- Dead token imputation: Delisted tokens without price data are imputed as total losses. This is a conservative assumption — some may have retained value on other exchanges.
- Not financial advice: This is an educational analysis. Past performance does not predict future results.