Thoughts before September 2026 Fed Decision

Following the hot CPI print with core CPI up 0.3% compared to the expected 0.2% MoM, the markets have drastically shifted in favor of a hike in September. I think I am in favor of a hike now, but the market puts it at 80% which I think are close enough to fair that I don’t think a good bet exists in either direction. All I am certain of is that a hike has to come in the future, subsequently the markets for hike by end of 2027 moved to 95% which is where I thought that they should be given the discussion of a hike this year and credibility.

It seems pretty hard to predict individual Fed decisions, but I guess that’s good because it makes it interesting. However, I do think there is edge in other markets that also may be easier or more fruitful than the individual decision markets. As with the idea I had above, I think that wasn’t too difficult of a conclusion to come to. My agent also found some interesting opportunities on the number of dissents at the meeting that actually ended up moving in the correct direction (however that could also just be luck).

I still am trying to figure out my style. One of the processes I’m working on is that certain markets are sort of primary drivers of others and you can analyze the narrative around the primary markets to come to conclusions about the other markets further down the chain. (So this would encompass the Fed decision in September -> Fed rate hike by end of 2027 idea). Another idea I have is analyzing upcoming catalysts that are determined by humans (Warsh Jackson Hole speech for example) and determining the directional move that the market will make after reacting to the catalyst happening. I think these are pretty predictable and perfect for a fundamental/discretionary strategy because they inherently can’t be predicted in a systematic manner.

Another observation I had is that it is really easy to look at prices and buy in at highs when most people or everyone believes a certain thing. I think that when the consensus agrees with you, it’s easy to get some conviction that is undue for your belief. And this can be true even if your reasoning is different from consensus. I think that’s probably why it’s common for retail traders to buy at highs and sell at lows: because the market really influences how they’re thinking in a way that messes with their convictions and beliefs. This is something I will have to be more concious of avoiding. Ideas need to be purely from common sense, the opinion of the market should not influence that reasoning. I do think it’s important to understand the market’s beliefs, but it’s a fine line to walk and something I will have to learn and develop to make sure I’m thinking independently.

In a weird way, I think sometimes the prediction is almost easier or more accurate when you don’t see the market’s price before you make it. I might have to try out stuff along those lines. I think seeing those prices or hearing opinions of others can potentially make you overthink and potentially muddle your own thinking.

Lastly, an update on my agent’s progress. Previously, my agent had prompts that outlined a process for each individual agent and mandated certain things that the agent get for a dashboard that were industry specific. I created a builder agent to help me make these prompts, but I realized this is inefficient. A lot of what I want each agent to have overlaps, the research process, the way I want them to do analysis and find ideas, etc. So I’m going a bit with the skill file approach right now where there will be these universal/modular prompt files that I always feed to any agent. I think a domain specific file is still necsesary, so the builder agent I had previously will need to be responsible for that and I will have to rework that. But this work has also made me try and think more deeply about what is the actual process for getting an idea. I think a decent amount of my own process is subconcious (as with any human), so codifying it in a prompt is difficult. I essentially am trying to put my philosophy in the agent which I learned over many years of lived experience, distilling that into a text file isn’t exactly easy especially for an LLM which needs extreme clarity in order to do what you actually want instead of misinterpreting what you wrote and doing something based on the misinterpretation.