Pith. sign in

REVIEW 4 major objections 6 minor 7 references

The Surprising Irrelevance of Total-Value-Locked on Cryptocurrency Returns

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read TVL-sorted crypto portfolios earn no abnormal return beyond the crypto market.

desk verdict A tidy but underpowered null: TVL-sorted portfolios show no alpha in a 105-week bull window, but the data cannot rule out a large premium. read the letter →

arxiv 2506.03287 v3 pith:LKUFIRDH submitted 2025-06-03 q-fin.PR econ.GNq-fin.EC

Matt Brigida This is my paper
classification q-fin.PRecon.GNq-fin.EC
keywords cryptocurrencytotalvaluelockedassetpricingfactormodelsDeFiportfolioreturnsmarketbetaalpha
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tests a common assumption in crypto markets: that total-value-locked (TVL) is positively related to a cryptocurrency's returns. It builds value-weighted portfolios sorted by TVL relative to market capitalization, and by changes in that ratio, over January 2023 through December 2024, and regresses their returns on a crypto market factor plus size and momentum factors. The paper's central claim is that all TVL-sorted portfolios with significant mean returns are spanned by the aggregate crypto market return: their alphas are statistically indistinguishable from zero, and a one-factor market model is sufficient. This matters because if true, TVL data cannot be used to form strategies that earn abnormal returns, and TVL-based strategies can be replicated simply by levered exposure to the crypto market portfolio.

What carries the argument

The central object is the TVL factor and the TVL quartile portfolios. Each week, assets are ranked by TVL divided by market capitalization in the previous week, and value-weighted long-short portfolios buy the top quartile and sell the bottom quartile; a second version uses the change in TVL-to-market-cap over two weeks. These portfolios' excess returns are regressed on the crypto market excess return, with a three-factor version adding a crypto small-minus-big factor and a momentum factor. The work this does is to separate market exposure from TVL-specific pricing: if the alpha is zero and the joint zero-alpha test does not reject, the TVL portfolios are spanned by the market factor and contain no independent information.

What would settle it

Re-run the same TVL sorting on data from the 2020-2022 DeFi boom-bust cycle (or any other out-of-sample window) and test whether the high-minus-low TVL portfolio shows a statistically significant alpha relative to the crypto market return; a significant alpha in such a window would contradict the paper's claim that TVL performance is fully explained by the market.

Watch

Extended reading notes

Core claim

The paper constructs TVL high-minus-low and quartile portfolios from both total TVL and a simple TVL measure that strips out staking, pool2, governance tokens, borrows, double counting, liquid staking, and vesting, plus a Level-1-token subset. For every portfolio with a statistically significant mean return, the estimated alpha from a crypto market model is statistically indistinguishable from zero, and the standard joint test of alphas being zero cannot reject the hypothesis across portfolios. Market betas are positive and typically above one, so high-TVL portfolios behave like levered positions in the crypto market. Adding small-minus-big and momentum factors changes the results only marginally, leading to the conclusion that a single-factor model using the aggregate crypto market return fully explains the cross-section of TVL-sorted portfolio returns.

Load-bearing premise

The 105-week sample from January 2023 through December 2024 is representative and long enough to reveal any return premium earned by TVL-sorted portfolios if one exists.

Editorial extensions

If this is right

  • TVL-sorted long-short strategies cannot deliver abnormal returns after controlling for crypto market exposure, so TVL does not support an independent risk premium.
  • Any TVL-based portfolio can be replicated by a levered position in the crypto market portfolio, making TVL information redundant for pricing.
  • Cleaning TVL to remove overstatement (staking, borrowing, double counting) does not create alpha, so the null result is not an artifact of inflated TVL measures.
  • The Level-1-token subset behaves the same way, so the result is not driven by small, illiquid, or exotic coins.
  • High-TVL coins move with the market with betas above one, implying they are simply higher-volatility market bets rather than safer or higher-return assets.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the sample covers a calm, generally rising market; TVL-driven mispricing may be episodic and concentrated in DeFi boom-bust cycles such as 2020-2022, so a regime-conditional test is the natural next check.
  • Editorial inference: TVL could still contain information about protocol usage, fees, or fundamental value even if it does not predict returns; the paper's result is about pricing, not about TVL as an activity measure.
  • Editorial inference: if the market portfolio is the only factor needed, then TVL sorting adds no value to risk management either; a direct test would be whether TVL-sorted portfolio residuals contain no predictable component beyond the market.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper constructs value-weighted portfolios of 335 cryptocurrencies sorted by total-value-locked relative to market capitalization (and by changes in that ratio), using both total TVL and an adjusted TVL measure, over 105 weekly observations from January 2023 to December 2024. It estimates a single-factor crypto market model and a three-factor model (market, small-minus-big, momentum) on the quartile portfolios that have statistically significant mean returns, and reports insignificant alpha coefficients and GRS tests that do not reject joint zero alpha. The paper concludes that TVL-sorted portfolios are fully spanned by aggregate crypto market returns, so TVL carries no independent pricing information.

Significance. If the result held, it would be practically useful: DeFi-exposed portfolios could be replicated with levered positions in the crypto market portfolio, and TVL would not constitute a distinct risk factor. The paper has notable strengths: a transparent Fama-French style procedure, a survivorship-conscious universe of the top-100 cryptocurrencies, use of both total and adjusted TVL, portfolio-level analysis that reduces idiosyncratic noise, and GRS tests for joint alpha. However, the evidence is underpowered and applied only to selected portfolios, so the strong conclusion of 'fully explains the cross-section' is not supported by the reported tests.

major comments (4)
  1. [Section 2.2, Tables 1-8] The 105-week sample yields wide confidence intervals on alpha. Portfolio weekly volatilities in Tables 1-6 range from about 5.7 to 13.5 percentage points, so the standard error of an estimated weekly alpha is approximately 0.5-1.3 percentage points. A true weekly alpha of 0.5-1.0 percent (26-52 percent annualized) is therefore inside the 95 percent confidence intervals of the reported estimates; for example, Table 8 Quartile 1 reports alpha of 0.25 with p=0.20, which is consistent with economically large mispricing. The GRS p-values of 0.35-0.99 show only that the test lacked power, not that alpha is zero. The abstract's claim that portfolios 'exhibit returns that are linear functions of aggregate crypto market returns' and the conclusion that a single-factor model 'fully explains' the cross-section overstate what the data can support.
  2. [Section 3, Tables 8-13] The GRS tests are applied only to the quartile portfolios whose mean returns were significant in the descriptive statistics, not to the full set of TVL-sorted portfolios. For instance, Table 8 tests only Quartiles 1, 3, and 4, while Table 10 tests only Quartiles 2 and 4. The central claim about the cross-section of TVL-sorted portfolios requires a joint test across all four quartiles (and ideally the HML portfolio) simultaneously. The current procedure is conditional on in-sample selection of significant mean returns, which both weakens the test and leaves the full cross-section untested. Reporting GRS statistics for the complete set of TVL portfolios would directly address this.
  3. [Section 2.2, sample period] The sample runs from January 2, 2023 through December 31, 2024, a period of generally rising crypto prices that excludes the 2020-2022 DeFi boom-bust cycle. If TVL-related mispricing is episodic and concentrated in such cycles, the zero-alpha result is an artifact of regime selection. At minimum, the conclusions should be limited to the sample period studied, and the paper should either extend the sample or explicitly acknowledge that the findings cannot support a general statement of TVL irrelevance.
  4. [Table 7, Table 3] The small-minus-big factor in Table 7 has a weekly minimum of -206.53 percent, and the HML portfolio in Table 3 has a minimum of -61.71 percent. These extreme values suggest possible data errors or outliers that could materially affect factor regression estimates, particularly for the three-factor models where SMB is a regressor. The paper does not report robustness checks such as winsorization, exclusion of suspect observations, or verification of the underlying weekly return data. The authors should investigate these observations and show that the main alpha results are unchanged when outliers are handled differently.
minor comments (6)
  1. [Table 8 caption] The caption says portfolios are 'formed on ΔTVL/Market Cap,' but the text and table header describe portfolios formed on total TVL to market cap; this appears to be a typo and should be corrected.
  2. [Section 2.1] The text states 'This high-minus-low TVL portfolio is denoted as HM Lin tables below,' but the tables use 'HML'; please fix the notation.
  3. [Section 3] The phrase 'We do not reject the null for any of the ibid. tests' is informal and unclear; replace 'ibid.' with 'the GRS' or 'Gibbons-Ross-Shanken'.
  4. [Tables 1-6] The tables report p-values for means and coefficients but no standard errors; adding standard errors or confidence intervals would make the precision of alpha estimates more transparent.
  5. [Section 2.2, Tables 1 and 2] Table 1 reports 105 weekly observations while Table 2 reports 104; the one-observation difference should be explained, presumably due to the change computation requiring an additional lag.
  6. [General] A data availability statement and the authors' code (if any) would improve reproducibility; the current manuscript does not mention whether data or code will be shared.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: alpha is freely estimated and the zero-alpha result is an empirical null, not a construction artifact.

full rationale

The paper's central claim is that TVL-sorted portfolio returns have insignificant alphas when regressed on the aggregate crypto market factor. The alphas and betas are freely estimated by OLS, and the GRS test treats joint zero alpha as a null hypothesis to be rejected. Nothing in the portfolio construction or factor definition forces alpha to be zero; the market factor includes the test assets, but that is standard in factor-pricing tests and does not make the alpha estimate an identity. The statement that portfolios 'can be replicated with appropriate weights on the crypto market portfolio' is a restatement of the estimated zero-alpha result, not a fitted parameter relabeled as a prediction. There are no self-citations: all references are to external work (Fama-French, GRS, Liu-Tsyvinski-Wu, etc.), and none is authored by Brigida. The main weaknesses of the paper—105 weekly observations, a single bull-market regime, and GRS tests applied only to quartiles with significant in-sample mean returns—are statistical power and selection concerns, not circularity. The derivation is therefore self-contained with respect to the circularity definitions.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central test relies on standard statistical machinery and on domain assumptions about data quality and sample representativeness. Hand-chosen parameters are limited to conventional sorting and lookback choices. No new entities are introduced.

free parameters (3)
  • Portfolio sort breakpoints = top/bottom 25% (quartiles)
    Chosen by convention following Fama-French style, not fitted to produce the zero-alpha result.
  • Momentum lookback window = 5 weeks (t-5 to t-1)
    The cumulative return window is set to 5 weeks by convention; it is not estimated from the data.
  • Observation frequency = weekly
    Weekly returns are used; this choice determines the 105-observation sample and the statistical power of the tests.
assumptions (5)
  • domain assumption TVL-sorted portfolio returns are generated by a linear factor model with the specified crypto factors.
    This is the standard asset pricing spanning assumption. If an omitted factor is priced, the alpha estimates are biased.
  • domain assumption DefiLlama TVL data and the simple TVL exclusions correctly measure locked value.
    The paper removes staking, pool2, governance, borrows, double count, liquid staking, and vesting, but the accuracy of this adjustment is assumed.
  • domain assumption Top-100-at-any-point sample controls survivorship bias.
    Including coins that were in the top 100 at any point in 2023-2024 reduces survivorship bias but is still an ex-post selection rule.
  • domain assumption Bitcoin and stablecoin exclusions do not remove the relevant cross-section.
    Bitcoin has no meaningful TVL and dominates market returns; stablecoins have near-zero return variance. The exclusion is reasonable but affects the test universe.
  • standard math OLS and GRS test assumptions hold (independent, homoskedastic errors).
    The paper does not test for serial correlation or heteroskedasticity; with 105 weekly observations, these assumptions may be violated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Surprising Irrelevance of Total-Value-Locked on Cryptocurrency Returns." pith.science (2026). https://pith.science/paper/LKUFIRDH

@misc{pith2026250603287,
  author       = {Pith},
  title        = {Pith review of: The Surprising Irrelevance of Total-Value-Locked on Cryptocurrency Returns},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LKUFIRDH}},
  note         = {Machine review of arXiv:2506.03287}
}
read the original abstract

A common assumption in cryptocurrency markets is a positive relationship between total-value-locked (TVL) and cryptocurrency returns. To test this hypothesis we examine whether the returns of TVL-sorted portfolios can be explained by common cryptocurrency factors. We find evidence that portfolios formed on TVL exhibit returns that are linear functions of aggregate crypto market returns, that is they can be replicated with appropriate weights on the crypto market portfolio. Thus, strategies based on TVL can be priced with standard asset pricing tools. This result holds true both for total TVL and a simple TVL measure that removes a number of ways TVL may be overstated.

Figures

Figures reproduced from arXiv: 2506.03287 by the authors.

Figure 1
Figure 1. Total TVL to Market Capitalization 2023-01 2023-03 2023-05 2023-07 2023-09 2023-11 2024-01 2024-03 2024-05 2024-07 2024-09 2024-11 2025-01 Date 0.035 0.040 0.045 0.050 0.055 0.060 Total TVL / Total MCap Ratio All Cryptocurrencies - Aggregated TVL/MCap Ratio All Cryptos 2023-01 2023-03 2023-05 2023-07 2023-09 2023-11 2024-01 2024-03 2024-05 2024-07 2024-09 2024-11 2025-01 Date 0.035 0.040 0.045 0.050 0.055 0.060 Tota… view at source ↗
Figure 2
Figure 2. Simple TVL to Market Capitalization 2.1 Crypto Factor Construction To construct the TVL factor we first scale TVL by market cap which af￾fords a proportion of market cap which is locked. Then, for a given week t, we rank assets by TVL / Market Cap in week t − 1. We then create a value-weighted long-short portfolio which buys the top 25% of coins by TVL and sells the bottom 25%. The return on this portfolio over week… view at source ↗
Figure 3
Figure 3. Cryptocurrency Returns by Total TVL to Market Capitalization [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Cryptocurrency Returns by Total TVL to Market Capitalization [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Cryptocurrency Returns by Simple TVL to Market Capitalization [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Cryptocurrency Returns by Simple TVL to Market Capitalization [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Level 1 Cryptocurrency Returns by TVL to Market Capitalization [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Level 1 Cryptocurrency Returns by TVL to Market Capitalization [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

7 extracted references · 3 canonical work pages

  1. [1]

    Multifactor Expla- nations of Asset Pricing Anomalies

    Fama, Eugene F. and Kenneth R. French (Mar. 1996). “Multifactor Expla- nations of Asset Pricing Anomalies”. In:The Journal of Finance51.1, 55–84. issn: 1540-6261. doi: 10.1111/j.1540-6261.1996.tb05202.x . url: http://dx.doi.org/10.1111/j.1540- 6261.1996.tb05202.x (cit. on pp. 2, 13)

    arXiv 1996

  2. [2]

    A TestoftheEfficiencyofaGivenPortfolio

    Gibbons, Michael R., Stephen A. Ross, and Jay Shanken (Sept. 1989). “A TestoftheEfficiencyofaGivenPortfolio”.In: Econometrica57.5,p.1121. issn: 0012-9682. doi: 10.2307/1913625. url: http://dx.doi.org/10. 2307/1913625 (cit. on pp. 13–19)

    doi:10.2307/1913625 1989

  3. [3]

    Trust as a Driver in the DeFi Mar- ket: Leveraging TVL/MCAP Bands as Confidence Indicators to Antici- pate Price Movements

    Grande, Mar and J. Borondo (2025). “Trust as a Driver in the DeFi Mar- ket: Leveraging TVL/MCAP Bands as Confidence Indicators to Antici- pate Price Movements”. In:Finance Research Letters. doi: 10.1016/j. frl.2024.104123. url: https://www.sciencedirect.com/science/ article/pii/S1544612324017343 (cit. on p. 3)

    arXiv 2025

  4. [4]

    Common risk factors in cryptocurrency

    Liu, Yukun, Aleh Tsyvinski, and Xi Wu (2022). “Common risk factors in cryptocurrency”. In:The Journal of Finance77.2, pp. 1133–1177 (cit. on p. 2)

    work page 2022

  5. [5]

    Piercing the veil of TVL: Defi reappraised

    Luo, Yichen et al. (2024). “Piercing the veil of TVL: Defi reappraised”. In: arXiv preprint arXiv:2404.11745(cit. on p. 2)

    arXiv 2024

  6. [6]

    Evaluating the Sig- nificance of the Total Value Locked to Market Capitalization Ratio

    Pantelidis, Konstantinos and Ioannis Karakostas (2024). “Evaluating the Sig- nificance of the Total Value Locked to Market Capitalization Ratio”. In: International Journal of Economics and Finance16.11, pp. 41–41.doi: 10 . 5539 / ijef . v16n11p41. url: https : / / ccsenet . org / journal / index.php/ijef/article/view/0/50786 (cit. on p. 3)

    work page 2024

  7. [7]

    Saggese, Pietro et al. (2025). Towards Verifiability of Total Value Locked (TVL) in Decentralized Finance. Tech. rep. 1268. Bank for International Settlements. url: https://www.bis.org/publ/work1268.pdf (cit. on p. 2). 21

    work page 2025

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.