To summarize, we have by now defined the following set of three orthogonal, but interconnected, tasks that constitute the minimal software engineering efforts required to build the MVP of our business model:
- Assemble an initial "minimum strategy set" using only known strategy types for the sake of simplicity;
- Find an easy, practical way to fit a subset of the chosen strategies to the evolved market when it comes time to deploy (i.e., what else can we do if not "strategy-market fit"?); and
- Apply proven machine learning techniques to build a "Type 2 Bot" that knows how to search and capture novel trades along the opportunity frontier.
But how does the completed MVP actually get to the opportunity frontier to conduct search and capture? MVP will need two things: a means of transport, and a good map.
We have found a fast ship that can transport MVP across the oceans to get to the opportunity frontier at reasonably low latencies. Not the lowest, but low enough. This works nicely for us as we are treasure hunters, not ship builders. The ship operator seems honest and reliable, and does not appear to have a hidden agenda (no, we don't think MVP is being sold into indentured servitude). We think an economy hammock on this ship may do quite nicely for now, as MVP has quite a few more of such trips planned ahead (if he survives this very first one!).
Next, we will need to give MVP a good treasure map, For this we enlist the help of a cartographer (aka "map maker"). She is pretty good at her craft; and an astute businesswoman, too. The cartographer gently reminds us that: "Finding the way once is luck. Finding the way twice takes a good map." We wholeheartedly agree. Lady Luck, who practices divination with a crystal ball just two doors down the hallway, appears to have an uneven temperament. Many of her trading clientele were known to have died mysterious deaths and were never heard from again.
"A diamond for each map," beams the cartographer, "but upon your return." She certainly has our best interest in mind and we appreciates her vote of confidence in our untested ability to bring back treasure. There is no doubt we'll be doing a few more such map-for-diamond trades with the cartographer; we just need to be sure our business stays afloat after this very first perilous journey at sea. We hope that we are not her guinea pigs. After all, she still holds her day job at the cartographic department at a local university.
While the details remain to be worked out and iterated over, we can't emphasize enough how important it is at this early stage: (i) to keep things simple, and (ii) to use only known methods or proven techniques; so as to minimize engineering costs and implementation risks.
Further improvements or enhancements can always be added on later, after first seeing how the MVP actually works in backtesting and/or on live data during paper trading. We envision the fifth generation MVP will be artificially intelligent with fully formed anthropomorphic interfaces, and a sight to behold.
Time for us now to get back to the lab. We've got ourselves an MVP to build!