There are several ways to work with BTC as digital capital. One approach is accumulation. This strategy is focused on increasing the amount of BTC over time. Another approach is cash flow. In this case, BTC is treated as productive capital that should help generate spendable income while preserving the BTC position. A hybrid strategy combines these two approaches: it tries to grow BTC-denominated capital over time while also generating some cash flow.
Regardless of whether an investor pursues accumulation, cash flow, or a hybrid strategy, effective DeFi tools are needed. One possible category of such tools is decentralized exchanges, or DEXs.
In this article, however, we are not discussing spot or perpetual trading as trading strategies. Instead, we are focusing on liquidity provision. A liquidity provider supplies capital to a trading venue and becomes exposed to the economics of that venue: trading fees, borrowing fees, trader PnL, asset exposure, and pool mechanics.
We will use GMX as a practical example. GMX is one of the longer-running and well-established decentralized perpetual exchanges. The focus of this article is not to promote GMX or to compare it with other DEXs. The goal is to understand the mechanics of liquidity provision in a perpetual DEX environment using a practical example supported by historical on-chain data.
Why Focus on GM Markets?
According to the GMX documentation, liquidity can be provided through both GM markets and GLV vaults. GM markets represent individual markets, while GLV vaults allocate liquidity across supported GM markets and aggregate their value.
The advantage of individual GM markets is that they allow the investor to choose a specific exposure depending on the investment objective. For example, as an investor focused on BTC as digital capital, I am not interested in broad exposure to markets such as ETH, SOL, LINK, or other non-BTC assets.
For an investor who prefers aggregated exposure across several GM markets, GLV vaults may be a suitable instrument. They can be useful when the goal is to delegate allocation across multiple markets. Similarly, an investor who wants to focus on another asset, such as SOL, can choose a specific SOL-related GM market and analyze it separately from the performance of other pools.
In this article, the focus is BTC as digital capital. Therefore, we will analyze one isolated GM market: GM: BTC/USD [BTC-USDC].
Time Is Our Friend
Before we move to the mechanics of GM pools and the components of their performance, one point should be made.
GM pools should not be evaluated as short-term yield instruments. They should be analyzed over a longer time horizon. Based on the behavior of these pools, a period of one year or more is more meaningful, while a horizon of up to two years gives a better framework for analysis.
The reason is simple: GM pool performance is not linear. It depends on fees, market direction, volatility, trader PnL, and GM pool mechanics. Some periods may look weak or even negative, while longer periods can show whether the pool is actually creating value.
The goal is not to outsmart the market over a few weeks. The goal is to understand the pool mechanics, evaluate the sources of performance, and decide when the instrument makes sense.
If you are looking for a quick way to make profit, this article is probably not for you. The approach discussed here requires patience, data, and a longer investment horizon.
What Is a GM Pool?
GM pools are the foundation of GMX’s trading infrastructure. Each market supported by GMX has its own isolated liquidity pool. Trading activity in a specific market is routed against the corresponding pool.
When investors provide liquidity to such a market, they receive GM tokens representing a share of that pool. This gives them exposure to the pool’s underlying assets and to fees generated from trading, liquidations, borrowing, and swaps.
The result of a GM pool position can be understood on two levels.
The first level is asset exposure. For example, GM: BTC/USD [BTC-USDC] has exposure to BTC and USDC. If BTC increases in price, the BTC exposure of the pool also increases in value. If BTC decreases in price, this part of the exposure decreases accordingly. Therefore, as long as the investor holds GM tokens, the position is affected by the price movement of BTC.
The second level is pool performance relative to the benchmark. This is the part of performance that we will analyze in the following sections. It can be understood as a combination of two components:
- Fee Component
- Non-Fee Component
The Fee Component refers to explicit fee income generated by trading, borrowing, swaps, and liquidations. In the data, this component is represented by Fee APR. Fee income itself cannot be negative. However, if trading activity decreases significantly during certain periods, fee income is also expected to decrease.
The Non-Fee Component refers to the part of pool performance driven by trader PnL and by GM pool mechanics outside explicit fee income. These mechanics may include price-impact effects, pool imbalance, changes in liquidity composition, and changes in long/short positioning.
If traders lose money, the pool benefits. Negative trader PnL increases the value of the pool for liquidity providers. However, if traders are profitable, the pool has to pay out their profits, which can reduce the performance of the pool.
In the following analysis, we will separate these two components. First, we will look at the Fee Component. Then, we will estimate the Non-Fee Component as the residual part of performance that remains after subtracting the Fee Component.
Fee Component
The next chart shows the daily Fee APR for the GM: BTC/USD [BTC-USDC] pool since the creation of the pool. The data was obtained from GMX on-chain data and corresponds to the series shown on the GMX website.
The raw daily data is noisy and contains sharp day-to-day spikes. For better readability, the chart also includes a smoothed version of the same series. The smoothing was applied using a centered moving average, which avoids a visual phase shift in the curve.
As expected, the Fee APR remains positive or approaches zero, but it does not become negative. This is because fees represent income generated by market activity. A noticeable decline in Fee APR can be observed from September 2025, around the period when BTC started to weaken. However, despite the weaker BTC market environment, the Fee APR still remained meaningful from a long-term liquidity-provider perspective.
The next two charts show the Fee Component from the perspective of different investment horizons.
Each curve represents a different holding period. For example, consider the 90-day curve. Each point on this curve shows the annualized Fee APR that would have been realized if an investor had entered the pool 90 days before that date and exited on that date.
For example, if the 90-day curve shows 20%, this does not mean that the investor earned 20% during those 90 days. It means that the fee return accumulated during that 90-day holding period corresponds to approximately 20% annualized, or roughly 20% × 90 / 365 for the actual holding period.
The 30-day curve is much more volatile. Depending on the entry and exit date, an investor could have realized a Fee APR close to 35%, but could also have realized a much lower result, for example around 5%. Longer holding periods, such as 180 or 365 days, reduce this volatility and make the realized Fee APR less dependent on the exact entry and exit date.
This illustrates how strongly the fee outcome depends on the length of the investment period and on the timing of entry and exit.
The next chart uses the same principle, but applies it to longer investment horizons.
As discussed earlier, time is our friend. A longer holding period reduces the volatility of the realized Fee APR and makes the Fee Component less dependent on short-term entry and exit timing. Even in a weaker BTC market environment, the Fee Component can remain meaningful for a long-term investor.
Non-Fee Component
The Non-Fee Component is driven by trader PnL and by GM pool mechanics. Unlike the Fee Component, it can work in both directions. During some periods, it supports pool performance. During other periods, it works against the pool.
Let us now estimate the Non-Fee Component and look at its volatility.
The GMX website shows the cumulative performance of a GM market over different time periods. The cumulative performance curve of the GM: BTC/USD [BTC-USDC] pool is shown below to make clear that the following analysis is based on real GMX market data derived from on-chain activity.
First, we calculate the annualized daily performance rate from the cumulative performance curve. This performance rate includes both the Fee Component and the Non-Fee Component.
Annualized Daily Performance Rate = (Today’s Cumulative Performance - Yesterday’s Cumulative Performance) × 365
The reason for annualizing the daily change is that the Fee Component is already expressed as Fee APR, which is an annualized value. To compare both components directly, they need to be expressed in the same format.
After that, we subtract the daily Fee APR from the annualized daily performance rate.
Non-Fee Component = Annualized Daily Performance Rate - Fee APR
The result is the annualized Non-Fee Component, shown below.
The next chart zooms in on the Non-Fee Component curve for better readability. This makes the volatility of the component much easier to observe.
The curve can move both above and below zero. When the Non-Fee Component is above zero, non-fee effects contributed positively to pool performance. This may indicate that traders lost money to the pool, but it may also reflect favorable GM pool mechanics.
When the curve falls below zero, the opposite is true. In these periods, non-fee effects reduced pool performance. This may indicate that traders were profitable against the pool, or that GM pool mechanics had a negative impact.
Now let us look at the same idea from the perspective of different investment horizons, as we did earlier for the Fee Component. The next chart shows the rolling holding-period Non-Fee Component for investment horizons between 30 and 365 days.
The interpretation is the same as before: each point shows the annualized Non-Fee Component that would have been experienced over the selected holding period. A negative value does not mean that the entire GM position was negative; it only means that non-fee effects worked against the pool during that period.
This view makes the timing risk much more visible. Imagine that an investor’s working horizon for a GM pool is only four to five months and that the investor is not prepared to wait longer. In such a case, the timing of entry becomes very important. For example, opening a position around January 2025 and closing it roughly in April 2025 would have exposed the investor to a clearly negative annualized Non-Fee Component.
This does not necessarily mean that the total GM position would have been negative, because the Fee Component and asset exposure also matter. However, it shows that the Non-Fee Component can work strongly against the pool over shorter holding periods.
The picture changes when the investment horizon becomes longer. The next chart applies the same rolling holding-period approach to horizons from one year to two years.
Here we can see that longer holding periods reduce the dependence on short-term entry and exit timing. Even when the Non-Fee Component remains slightly negative, the overall GM position can still remain meaningful if the Fee Component is strong enough.
The recent period should be interpreted carefully. In the data shown here, the market environment has been strongly bearish, and this may explain why the Non-Fee Component is still under pressure. If market conditions change, the Non-Fee Component may move back toward zero or even become positive again. However, this should be treated as a scenario, not as a guarantee.
The main conclusion is that GM pools require a longer investment horizon. Sometimes the better decision may be not to close the position too early, but to wait until short-term negative non-fee effects have had time to normalize.
Putting the Components Together
Now we can bring both components together and look at the total GM pool performance rate.
The total annualized performance rate can be understood as the sum of the Fee Component and the Non-Fee Component:
Total Annualized Performance Rate = Fee Component + Non-Fee Component
In the data, the Fee Component is represented by Fee APR, while the Non-Fee Component is estimated as the residual part of performance after subtracting the Fee Component.
This chart shows that the Non-Fee Component can fluctuate strongly in both directions. It can move far above zero and support pool performance, but it can also move below zero and work against the pool. As a result, the Non-Fee Component can have a significant influence on the total annualized performance rate.
The Fee Component, on the other hand, acts as a positive base layer. It supports the performance rate because fee income is generated by trading, borrowing, swaps, and liquidations. However, the final performance rate is not determined by fees alone. It is the combination of the relatively stable Fee Component and the much more volatile Non-Fee Component.
The key point is that GM pool performance should not be understood simply as “earning APR from fees.” Fees are only one part of the result. The other part comes from trader PnL and GM pool mechanics, and this part can materially change the outcome over shorter and medium-term holding periods.
Why does the Non-Fee Component show such strong movements in both directions? We will look at this question in the next section. This will help us better understand perpetual DEX liquidity mechanics in general, and GM pools in particular.
Market Sentiment and BTC Price
In reality, nobody knows with certainty where the market will move next. In social media, news channels, and expert commentary, it is common to hear completely opposite opinions at the same time.
What we can observe, however, is market sentiment. During some periods, market participants are guided less by calm analysis and more by fear, greed, or crowd behavior. When the general sentiment is bearish, traders may become more willing to open short positions or avoid long exposure. When the general sentiment is bullish, long positioning may become more attractive.
However, markets are more complex than sentiment alone. Liquidity traps, short squeezes, long squeezes, and sudden reversals can occur at any time. A rising market can first collect liquidity below key levels before continuing upward, and a falling market can temporarily move upward before continuing lower. This is one reason why the Non-Fee Component can be difficult to predict in the short term.
Should an investor be afraid of the Non-Fee Component and avoid GM pools completely? This is a question every investor has to answer individually. However, historical data can help us understand how the components behaved under different market conditions.
To make this more visible, let us now compare three curves on the same chart:
- BTC price movement
- Fee Component
- Non-Fee Component
When reading this chart, it is important to remember that BTC price is shown on the right axis, while the Fee Component and the Non-Fee Component are shown on the left axis. Therefore, the vertical position of the BTC line should not be compared directly with the component lines. What matters is the market phase: whether BTC was rising, falling, correcting, or moving sideways.
The chart does not prove direct causality between BTC price movements and the Non-Fee Component. It only helps us observe how the components behaved during different BTC market phases.
The first important phase appears between late 2023 and spring 2024. During this period, BTC moved strongly upward. At the same time, the Non-Fee Component moved from negative territory into positive territory. This suggests that non-fee effects were working in favor of the pool. One possible interpretation is that some traders were positioned against the market move, or that they did not expect such strong bullish behavior from BTC.
The second phase appears around autumn 2024. BTC entered another strong upward move, and the Non-Fee Component increased sharply into strongly positive territory. The pool did not simply benefit from the Fee Component during this period. The Non-Fee Component also became a major positive contributor to the total performance rate. This may indicate that traders, on average, were again not positioned well against the market move, or that GM pool mechanics worked in favor of LPs.
The third phase can be seen between late 2024 and February 2025. After the previous strong BTC move, the market entered a weaker or corrective phase, while the Non-Fee Component moved sharply negative. In this period, non-fee effects worked against the pool. This may indicate that traders were better positioned against the pool, or that GM pool mechanics became less favorable for liquidity providers.
Another important phase appears around February and March 2025. BTC started to rise again, and the Non-Fee Component moved strongly back into positive territory. This shows how quickly the Non-Fee Component can change direction. During this phase, the pool received strong support from non-fee effects. The Fee Component remained the positive base layer, but the main movement in the total performance rate came from the Non-Fee Component.
The fifth phase begins around mid-2025 and continues into autumn 2025. During this period, BTC first moved higher and then started to weaken significantly. The Non-Fee Component eventually moved deeply into negative territory. This period is especially useful for understanding the downside risk of the Non-Fee Component. It shows that during a strong bearish market phase, the Non-Fee Component can become a serious headwind for liquidity providers. A possible explanation is that traders were positioned correctly for the downside move, or that GM pool mechanics worked against LPs during this period.
The sixth phase starts around the beginning of 2026. BTC remained weak or volatile, and the Non-Fee Component stayed mostly negative. This shows why short and medium-term holding periods can be difficult in GM pools. Even if the Fee Component remains positive, the Non-Fee Component can dominate the result over shorter periods and reduce the total performance rate.
The final phase in this chart appears in spring 2026. BTC partially recovered, and the Non-Fee Component moved back into positive territory. This illustrates that the Non-Fee Component should not be understood only as a risk. It is a two-sided force. During some periods, it works against liquidity providers. During other periods, it can become one of the strongest contributors to pool performance.
Overall, the chart shows that the market moves through distinct phases. Sometimes non-fee effects support the pool, and sometimes they work against it. The Fee Component behaves more like a relatively stable positive base layer, while the Non-Fee Component is much more sensitive to market phases, trader positioning, and GM pool mechanics.
In the current data set, the combination of the Fee Component and the Non-Fee Component has produced a positive annualized performance rate. However, this should not be interpreted as a guarantee for the future. For an investor who treats BTC as digital capital, the main value of this analysis is not prediction. The value lies in understanding how GM pool performance behaved historically during different BTC market phases and how strongly the final result can depend on the Non-Fee Component.
A Note on DeFi Risk
DeFi instruments are exposed not only to market risk, but also to technical and operational risk. These may include smart contract vulnerabilities, oracle failures, liquidity shocks, governance decisions, and infrastructure problems. In extreme cases, such risks may lead to a permanent loss of capital.
For this reason, it is good practice to study the history of a DeFi instrument before interacting with it. This may include checking whether the protocol has experienced past exploits, how those incidents were handled, how the protocol recovered, and how affected users or liquidity providers were treated.
It is also important to check how the protocol tries to reduce these risks through rigorous testing, independent audits, and active bug bounty programs. Mature protocols often provide this information publicly in their official documentation.
However, audits and bug bounty programs do not eliminate risk. They only provide additional information about the protocol’s security process. Historical performance data can help us understand how a pool behaved in the past, but it cannot remove technical or operational risk.
Conclusion and Outlook
In this article, we verified that the data obtained for the GM: BTC/USD [BTC-USDC] pool is consistent with the data displayed on the GMX website. Using this data, we separated GM pool performance into two components: the Fee Component and the Non-Fee Component.
The analysis shows that working with GM pools requires a longer investment horizon. In the short term, the result can be strongly affected by market phases, trader positioning, and GM pool mechanics. An investor who closes a position too early may miss a period of recovery or may realize a weaker result during a temporary negative phase of the Non-Fee Component.
The analysis also shows that GM pools should not be viewed only as simple fee-income products. The Fee Component provides a relatively stable positive base layer, but the final performance depends on the interaction between fees, trader PnL, asset exposure, and pool mechanics.
This makes GM pools relevant not only for investors working with stablecoins, but also for investors who treat BTC as digital capital. Depending on the strategy, GM pools may be used in different ways: for generating stablecoin-denominated cash flow, for accumulating BTC, or for building a hybrid strategy that combines both objectives.
For example, an investor may use borrowed stablecoins together with BTC exposure, or may analyze GM pools as a tool for accumulating more BTC during weaker market phases and generating stablecoin-denominated cash flow during stronger market phases. These strategies require careful simulation, because the result depends on market direction, holding period, borrowing costs, and the behavior of the Non-Fee Component.
In the next article, we will analyze these strategies in more detail and test them using simulations. Later, we will also examine the GM: BTC/USD [BTC-BTC] pool separately and discuss how it may be used by investors who want to work more directly with BTC-denominated capital.
The main conclusion of this article is simple: GM pools are not passive yield products in the traditional sense. They are dynamic liquidity instruments. Their performance comes from several sources, and understanding these sources is essential before using them as part of a long-term BTC strategy.