Adaptive Portfolio Strategy
Last updated
Was this helpful?
Last updated
Was this helpful?
In pursuit of mastering concentrated liquidity management (CLM), Teahouseās Strategy Team developed the Adaptive Portfolio Strategy. This article shares the design concept, backtesting results, and the final selected candidate of the Adaptive Portfolio Strategy.
Our approached LP provision with an inventory management mindset. This time, Teahouseās Strategy Team explored various established portfolio-asset-selection strategies in active CLM to see which has the best historical performance.
No one portfolio strategy can trump all in every market condition. Quite often (and in the case of the wETH/wBTC pair we tested), the simplest āoldie but goodieā strategies work the best.
PnL is influenced by the exchange rate of the tokens in the token pair, and a stronger token will show a lower ROI in a given period, all else being equal.
Backtesting shows a +23% APR (in wETH) and a +18% APR (in wBTC) during the 360 days tested.
To build the Adaptive Portfolio Strategy, Teahouseās Strategy Team selected various established portfolio strategies to see which performed the best in recent times. The portfolio strategy that yielded the best results was adopted and subjected to regular reassessments along with the other candidates to ensure that we consistently provide the top-performing strategy to our users. This reassessment process makes the strategy āadaptiveā to current market conditions.
The overall liquidity is divided into multiple positions and distributed on both sides of the spot price. Each portfolio strategy will have a different position weighting, with the goal of optimizing the allocation of assets within a liquidity pool. (i.e. each position A ā G will have various liquidity depths assigned to their price range)
As the spot price shifts within the position stacks, the swap fees earned will reflect the ROI of each position, which will serve as our metric for evaluating each portfolio. The results of this evaluation will then be dynamically used by each strategy to adjust the weights of the various positions.
Teahouseās Strategy Team evaluated six well-known, āestablishedā strategies along with Teahouseās automated LP strategy using different parameters.
The six established strategies can be categorized into three groups:
Balanced Approach strategies aim to maintain a stable and equal weight for all positions within the portfolio.
Constant Rebalanced Portfolio (CRP): This method involves maintaining an equal weight for all positions in the portfolio. Over time, as the prices of assets change, the portfolio is periodically rebalanced to restore equal weights.
Dynamic Constant Rebalanced Portfolio (Dynamic CRP): Similar to CRP, this strategy also rebalances periodically. But instead of maintaining an equal weight across all positions, it uses historical data to obtain an optimized weight for each position.
These strategies determine the weight allocation of assets based on their historical performance. They allocate higher weights to assets that have shown strong recent performance, often called āwinnersā. The aim is to ride the momentum of assets currently performing strongly in hopes of maximizing returns.
3. Universal Portfolio (UP): The UP strategy emphasizes the āwinnersā: assets with better recent returns receive higher weightings, while those with poorer recent performance are allocated less weight. UP applies a linear rate to adjust position weights.
4. Exponential Gradient (EG): EG is a strategy that strongly prioritizes the āwinnersā, and employs an exponential rate to modify position weights. This means the algorithm assigns greater weight to assets with strong performance, with a compounding effect.
These strategies do the opposite of momentum-based strategies as they aim to capitalize on profit opportunities in trades that go against the current market sentiment. They allocate higher weights to assets that have recently performed poorly, often called ālosers,ā with the expectation that they may revert to their historical mean.
5. Passive Aggressive Mean Reversion (PAMR): Unlike the previous strategies, PAMR roots for the ācomeback kidā, as it allocates higher weights to assets that have been underperforming or have had poor recent performance. The strategy assumes that these underperforming positions will likely revert to their historical mean, meaning they may recover and perform better in the future.
6. On-Line Portfolio Selection with Moving Average Reversion (OLMAR): OLMAR is a ālooking for reboundā algorithm that uses the historical moving average of asset returns to determine when to rebalance the portfolio. This method aims to benefit from the mean reversion behavior in asset prices (rebounds) by allocating a higher weight on positions with performance that are below the moving average.
Pair tested: wBTC-wETH (on Arbitrum)
Rebalance schedule: Fixed rebalance timeframes of every 6 hours, 12 hours, 1 day, and 2 days.
Backtesting period: 180 days or 360 days between 2021/9/12 and 2023/5/11.
Baseline (āBasicā in the chart Legend): Maintaining a portfolio comprising 33% wBTC, 33% wETH, and 33% in the 50:50 liquidity pool with a broad price range covering all backtest periods (either 180 days or 360 days).
Additional information that can help you read the charts below:
Strategies labeled with LP_X%(Y) are Teahouseās own automated LP strategy with different parameters. For example, the LP_80%(30) strategy sets the price range to cover 80% of the historical price in the 0.03% fee tier pool.
Strategies labeled with LP_single(Y) deploys liquidity in a single narrow position that follows the spot price, in the given Y fee-tier pool.
Among the three categories of Portfolio Strategy approaches, the āBalancedā approach generally yielded better returns in our pair.
In the shorter 180-day timeframe, the CRP strategy outperformed some of Teahouseās own automated LP strategies with basic parameter settings in terms of ROI.
The top-performing candidates showed an overall +23% APR (in wETH) and +18% APR (in wBTC) during the 360 days tested.
As PnL is influenced by the exchange rate of the tokens in the token pair, a stronger token will show a lower ROI in a given period. (i.e. In our tests, wETH had a higher APR than wBTC, indicating that wBTC was stronger during the period tested)
Of all the strategies assessed, the most fundamental approach of providing equal weights yielded the best results. (heh)
Despite showing a higher accumulated profit in the 0.3% fee tier during the backtests, given the recent shift in swap traffic heading to the 0.05% fee tier, deploying the strategy on the 0.05% fee tier pool would have been more capital-efficient.
Returning to the concept design for the Adaptive Strategy, our team initially theorized that a more āestablishedā portfolio approach might produce better results. However, our findings show that no single portfolio strategy outperforms all others. In contrast, and most interestingly, the most basic 50:50 equal-weight strategy (50% in WETH, 50% in WBTC, with equal weights across all positions) consistently performed the best over the time period tested.
Therefore, at launch, the Adaptive Strategy initially adopted the 50:50 equal-weight strategy and Teahouse Team will regularly evaluate other portfolio strategy options going forward.
