Proposed Vote: New Policy to Increase XAI Demand / Search for More Effective Trading Models

We’ve all been watching Wintermute’s performance over the past 6 months with some disappointment. The models have been very slow to take on board new information and have not been able to identify a profitable trading style in a market with much-reduced volatility.

Part of this is the reality of trading in a relatively high cost market with only small profit opportunities eating up capital in slippage; but we are also struggling with the history of crypto markets vs current reality. The models have deemed that trading in a manner that suits current market conditions would not be profitable – long term.

This dry spell has led to a number of people in the collective suggesting how we can improve the returns or generate returns from an alternative source. One idea several people have suggested is finding some way to utilise the collective to assist with active trading (not just selecting investments for the start-up pool).

Specifically, a number of people have expressed interest in feeding other models into trading decisions (either as an input into Wintermute’s decisions or as a separate collective driven model pool).

It is possible that there are other models that may be more effective trading in the current market. However, for this to work some thought would have to go into putting this into a manageable framework. Also, since this is a change from the current strategy it would have to be put to the collective for a vote.

The strawman model I would like to propose is as follows:

1. We will split out a sub-pool from the Wintermute Coin Pool. (The amount of this pool to be decided by vote). This will reduce the allocation that Wintermute can trade by the selected amount.
2. We will provide an API where the collective can connect their models to trade any of the markets we are currently set up to trade.
3. For the first month, the API (automated way to place orders) will connect to a “paper-trading” account.
4. If a model is profitable in month 1 will be used in month 2 for live trading. Decisions on trading coins will be a pro-rated position based upon all the models that successfully entered live trading in a given month. (As time goes on, higher weightings will be awarded to continuously profitable models.)
5. Each month the same process will be followed with models being shuffled in and out on the first Friday of every month based on their previous month’s performance.
a. Paper Trade models that show a month of profitable trading will be adopted.
b. Live models that show no profit or a loss in the month will be dropped.
6. During the month if the Collective model sub-pool makes a new equity high then the participants who provided the live models will share 5% of the new equity high profit.

The technical team will develop the API that allows a model to trade a number between +100% (full long position) and -100% (full negative or short position). So depending on the strength of the trading signal, a model will be full or partially committed to its decision. What this means in terms of capital allocated to the trade is based on the amount of capital allocated to a particular model and a particular market. If a collective member writes a model, and connects it to the AICoin API, they and the entire collective will be able to review the paper trading results. If the model is profitable and is then added to the models that are used to ‘live trade,’ that will happen automatically without any additional activity needed by the model developer.

To be permitted to add a model to the pool vying for live trading, the model developer must be a member of the collective holding at least 10,000 XAI (the XAI can be held on dashboard or in a designated off dashboard wallet.) The funds must be in the wallet on the first day of every month that the member’s model is trading. If the collective member sells their 10,000 XAI so it is not present in the designated wallet address at the start of the month, that member’s model will be automatically halted at the end of the month.

In addition to the direct benefit to the collective of the profits generated by this new crop of models it will also increase demand for XAI from model developers wanting to participate in the program. As an additional benefit, the decisions of these models will be added into the learning pool to potentially improve the Wintermute decision process.

As always, please send back any feedback on this idea so we can sculpt the decisions to put them to the vote (or discard if there is no interest in this approach).

On a general note, thank you to all the people who are continuing to come up with interesting ways we can try and generate additional returns during this difficult period. We’re looking at all the ideas proposed and seeing how we can structure them into proposals that can be put to the vote.

Go to the Voting page to read about the choices for this vote and the voting process.

Please post questions here so the whole collective can review them.