The first thing almost everyone types into a sports assistant is a version of who should I start? And the answer that comes back is a version of it depends — a paragraph of hedging with two names in it and no clear direction.
Then people conclude the assistant is useless and stop using it.
Which is a shame, because the same tool will happily walk you through the specific reason one of those two names has a genuinely different set of outcomes available to him this week than he did last week — the workload change, the matchup, the thing the beat writer said on Thursday that nobody picked up. It just needs to be asked something it can get hold of, and who should I start is not that.
It is not entirely the assistant's fault. That question, as asked, has no answerable form. It contains no criteria, no context, and no indication of what would make one answer better than another. A knowledgeable friend would respond by asking you three questions before saying anything, and the reason the assistant does not is that it is trying to be helpful immediately.
Why vague questions produce vague answers
Who should I start? is missing everything that determines the answer.
It does not say whether you are chasing a big week or protecting a lead. It does not say whether this is a league where you are already 7-2 or one where you need a miracle. It does not say whether you would rather have a reliable middling score or a coin flip between excellent and nothing.
Those are not details. They are the entire decision.
The same two players produce different correct answers depending on which of those is true, and without them the only honest response is the hedge you got.
A knowledgeable friend would ask you three questions before answering. The assistant does not, because it is trying to be helpful immediately.
Four ways to make a question answerable
Say what you are optimising for. "I need the highest ceiling, I am eleven points down with one player left" gets a real answer. "I need to not blow this, I am up thirty" gets the opposite real answer. Same matchup, same players.
Give it the constraint. "Which of these two, given I already have his quarterback" changes the shape of the problem. So does "I can only make one move this week."
Ask for the reasoning, not the verdict. "What is the case for each side" is a better question than "which one." You get the inputs and you make the call, which is both more useful and more honest about what the tool is actually good at.
Ask what would change your mind. This is the underused one. "What news between now and Sunday would flip this?" turns a static answer into something you can act on for three days.
The follow-up is where it gets good
Single questions get generic answers. The value shows up in the second and third exchange, which most people never reach because they were disappointed by the first.
Ask the broad question, get the hedged answer, then push. Why does the matchup favour him? What is the downside case? How much of this is the injury and how much is form? Each round narrows the thing and, more usefully, surfaces the assumption underneath — which is frequently where you find you disagree.
That disagreement is the point.
You are not looking for a verdict to obey. You are looking for the reasoning laid out clearly enough that you can find the part you think is wrong.
Ask where you already are. The RPO assistant sits on every screen and knows what you are looking at, so you can ask about the thing in front of you without describing it first.
Context you should not have to type
A fair objection: this is a lot of work for a question. If I have to specify my situation, my constraints, and my priorities every time, typing it out costs more than it saves.
Correct, and that is a design failure rather than a user problem.
An assistant that already knows which leagues you are in, which roster is yours, what your risk style is, and which screen you are currently looking at should not need most of that typed. The question you should have to ask is the short one; the context should already be there.
This is the difference between an assistant bolted onto a product and one built into it. A general chatbot needs you to explain your entire situation because it has no way of knowing any of it. One that lives inside the app you are already using has no such excuse.
You are not looking for a verdict to obey. You are looking for reasoning laid out clearly enough that you can find the part you think is wrong.
What it should not do
Two things worth being clear about, because the category has a credibility problem.
It should not pretend to certainty it does not have. A confident-sounding answer to a genuinely uncertain question is worse than a hedge, because you will act on it. Hedging when the situation is actually close is correct behaviour, not a failure — the failure is hedging when it is not close.
It should not hide the reasoning. An answer you cannot inspect is a tip, and tips are worth nothing because you cannot tell a good one from a lucky one. Anything worth acting on should come with enough of its own argument that you could reconstruct it and decide the assistant was wrong.
The short version
Stop asking who to start. Ask what the case is for each, given the situation you are actually in, and what would change it.
Then argue with the answer. That is the part where you get better, and it is the part a verdict would have skipped.
Ask it where you are
The RPO assistant is on every screen and already knows the context — your leagues, your rosters, your posture, and the thing you are currently looking at. Ask the short question and skip the setup.
Download RPO on iOS and push it past the first answer.
Related reading:
