RJ Mock One Discussion Thread

[DMT]

So rotolab has me 16th in Hrs but 1st in RBIs, that would be quite a feat. Also 11th in saves rather than 1st.

[revo]

Just glancing over those projections, the ERA and WHIP values look extremely optimistic. Worst ERA is 3.8? That would have placed 2nd in my 14 team mixed last year, 2nd in a 15 team mixed, and 4th in another 15 team mixed. And it's 17th?

Worst WHIP is 1.28 which seems very low as well.

Strange. I wonder why the projections are that far off of last year's numbers.

Like I said, it's a combo of BBHQ and my own projections. BBHQ is unusually optimistic, both in IP and ERA, which is why I tend to change them.

In addition, this takes only the starters, not any bench players into account.

[Ken]

Like I said, it's a combo of BBHQ and my own projections. BBHQ is unusually optimistic, both in IP and ERA, which is why I tend to change them.

Makes sense, good info to have. Thank you!

[revo]

Makes sense, good info to have. Thank you!

Just a for instance, it had Steven Brault with an ERA of 3.55. Where did that come from? Beats me. His previous two years he had ERAs of 4.86 & 4.67.

But OTOH, I don't like the projections to be too pessimistic, because then the player gets buried on the list. Another for instance, while Matt Harvey may indeed have an awful year, I'd rather not have projections for him in the 5.00+ range because then I'd never find him even though I like him as an end-game flier.

[cool hand]

Thanks revo! I feel much better about this projection than the one from Steamer. Like I said, I never did like Steamer much...
I'd have been shocked if I had a bottom tier offense.

[Ken]

Thanks revo! I feel much better about this projection than the one from Steamer. Like I said, I never did like Steamer much...
I'd have been shocked if I had a bottom tier offense.

Comparing projection systems is an interesting and relevant discussion point. Here's some resources from fantasypros:

https://www.fantasypros.com/2017/03/the-most-accurate-fantasy-baseball-projections-2016/

https://www.fantasypros.com/2016/03/most-accurate-baseball-projections/

https://www.fantasypros.com/mlb/accuracy/projections.php

Steamer performs well - in the top 3 every year.

Oddly BaseballHQ was not analyzed each year but in 2017 they were pretty bad relative to other projection systems, as far as accuracy (obviously the comparison algorithm is relevant and I have not tried to poke holes in that but I'm sure there may be some)

https://www.fantasypros.com/baseball-projections-accuracy/

This was another article I read last year on projection systems and their methodology and I found it very interesting:

https://www.beyondtheboxscore.com/2016/2/22/11079186/projections-marcel-pecota-zips-steamer-explained-guide-math-is-fun

[cool hand]

This is just my guess, without formal analysis other than spending 20 years looking at them, but it has always seemed like the sources touted as most accurate (the "deadly accurate" PECOTA springs to mind) are generally those with fairly conservative algorithms. They're stable and fair and it makes sense that, over the entire data population, they would be accurate. They seem to err on the side of caution on a consistent basis, and I understand why that is necessary. I think they do a pretty fair job of forecasting established players, and it seems likely that their deviation from actual results on those established players (the bulk of the league) is what drives those accuracy lists. Steamer has always reminded me a bit of the PECOTA forecasts, not in how they are generated but in their results, and I would never use either while building a team.

Roto leagues can be won & lost on breakouts and identifying players that deviate from expectations, for any of a number of reasons. A lot of time those "breakouts" are based on opportunity, among a litany of other factors, and these forecasts require a fair deal of "massaging", IMO. The strictly mechanical forecasts, while largely "accurate", can miss the boat on a number of other things that can separate winning from losing in roto.

Having said that, there is also something to be said for conservative forecasting being a strong backbone to winning in roto as well. Stability counts.

Anyway, I'm not expert at forecasting, or building algorithms to do so. I am, however, relatively successful in the league that I play in, and these are just my own theories and observations that have worked for me over the years. Sorry to be rambly, and extremely subjective, but I'm jotting this down as I think between bites at lunch.

[Ken]

I guess we use our projection systems very differently. I don't want the projection system to tell me what *could* happen.

I can figure that out myself. I want it to tell me what is most likely to happen based on the data set.

Then, if I want to apply my own ideas on top (which I always do), I can make my own exceptions and adjustments.

[cool hand]

I guess we use our projection systems very differently. I don't want the projection system to tell me what *could* happen.

I can figure that out myself. I want it to tell me what is most likely to happen based on the data set.

Then, if I want to apply my own ideas on top (which I always do), I can make my own exceptions and adjustments.

That's no different than what I'm saying. But what we're talking about here is using those projections, post draft, to project standings. Interpreting those results is quite different than how you use projections to draft.

[Ken]

Two different topics:

Topic 1) Should we use projections to look at projected standings?

My answer - with the full knowledge that it's far, far from perfect, and using it how it is intended - for example to highlight areas of weakness and strength, NOT to say team A is going to have 121 SB and team B will have 122 SB so due to my precise calculations team B is better - I think it is both useful and interesting.

But again, it's just analysis, it's not perfect. The better the projection system is at predicting that information, the better that analysis can be.

Topic 2) What projection system should we use for this analysis?

My answer: I posted some resources above but this is a very interesting question to me and I'm anxious to hear opinions. The resources I posted suggest that BaseballHQ, at least for 2017, was not very good at projecting accurate stats, relative to other systems. And historically Steamer has been good, again in a relative sense. Now that does not mean that I should look at Acuna and say just because Steamer projects X HRs I should pick him here. Rather, in the abstract general sense if I'm looking at a large data set (i.e. an entire league), it is likely the better resource to use. The total sum of all the data is closer to reality (and as I noted above, we see that in the ERA/WHIP projected by baseballHQ). At least based on the comparison's I've seen, Steamer is the better way to look at this. If there are more apt comparisons then I'd love to see them!

[Mithrandir]

Didnt something like the extremely live ball hr barrage lady year render any projections way off base? So what good were they?

[Ken]

Didnt something like the extremely live ball hr barrage lady year render any projections way off base? So what good were they?

Is your question whether 2017 projections were off due to the increase in HRs year? If so, I suspect they certainly were. Although that's likely the case for all projections (including the "what does my gut say" projections that individuals may make).

What "good" were they? I guess it's all relative. Everyone uses "projections", whether they are published projections or custom projections or just gut feelings. We don't draft Mike Trout because we like his weather predictions, we draft him because we "project" that he'll perform at a certain level, which is better than some other options performs at.

[Mithrandir]

Just to clarify: Lineups are set weekly. Are there any days that changes can be made or is the lineup for the entire week and that is it?

Thanks.

[BuckyBuckner]

Are we using FAAB to pick up players? Or do I now know how to use this site?

[joncarlos]

Are we using FAAB to pick up players? Or do I now know how to use this site?

I think we're using FAAB this year.