Psst... We're working on the next generation of Chowhound! View >
HOME > Chowhound > San Francisco Bay Area >
Jun 5, 2014 09:14 AM

The FiveThirtyEight Does Burritos

There's a great article on the 538 on the best burrito in America (Yes, I know, another poorly researched article displaying vast ignorance).
For those of you who aren't familiar with Nate Silver and the 538, he writes incredibly insightful articles on politics and sports (mostly) using very careful statistical analyses. In this article, he turns the same techniques on burritos.

  1. Click to Upload a photo (10 MB limit)
  1. Garbage in, garbage out. According to Silver's analysis of Yelp data, the four burrito-selling establishments in Northern California that rank in the top 20 nationally are all in San Francisco: El Farolito, La Taqueria, Taqueria CancĂșn, and HRD Coffee Shop.

    I tried an HRD spicy pork kimchi burrito. Not very good. Mediocre ingredients added up to less than the sum of its parts.

    6 Replies
    1. re: Robert Lauriston

      The Yelp ratings are there to guide the committee, but they are allowed to contribute ones not on the list.

      If they do stick to El Farolito, La Taqueria, Taqueria CancĂșn, and HRD Coffee Shop, San Francisco will likely lose. La Taqueria's limp burrito won't make it past the first round, and taqueria Cancun and Farolito will be undone by their inconsistency. HRD is good, but not as good as Cancun and Farolito at their finer moments.

      1. re: hyperbowler

        Ordered the carne asada burrito without beans from La Taquaria the other day. It costs $1.25 extra for no beans because they add more meat.

        That was easily the most tender, juicy and flavorful carne asada I have had from a taquaria in quite some time. The accompanying pico de gallo was also fresh, and the spicy salsa verde put the whole thing over the top.

        Everyone has there favorite, but for me La Taquaria is unbeatable.

        1. re: Civil Bear

          No beans would solve a bunch of problems. Their carne asada is awesome in the tacos, and is strongly flavored enough to compete with the beans in the regular burrito. I'll have to revisit it for full on meat.

          1. re: Civil Bear

            La Taqueria will grill a prepared burrito as a special request according to this article about their "secret menu":


          2. re: hyperbowler

            And, yelp is based in SF. We created that monster ourselves, we should be hoist by our own petard.

          3. re: Robert Lauriston

            Agree about the HRD burrito and most all of their stuff really, using super low quality and fatty meats. The kimchi burrito is just greasy and actually pretty gross.

            1. re: TVHilton

              Yup...good stuff. Mission burrito vs. LA burrito all over again:

              "so, what's the best way to troll California out of its goddamn mind?"

              1. re: ML8000

                Northern California & SoCal should be at least two of the 4 brackets. Maybe three, but I'm not sure how you'd split it up.

                1. re: TVHilton

                  I dunno, I suspect they're pretty good in Texas and New Mexico too. They at least deserve a peek.

                    1. re: dunstable

                      There is fantastic Mexican food in New Mexico (and I imagine Texas), but it's very different from what we know, and burritos aren't really part of it. So sure, there are probably some very good burritos there, but nothing like the culture of burritos that exists in California.

              2. Bill Addison and Gustavo Arellano are on the committee, so that could have added a lot of Bay Area credibility.

                But it didn't. The four SF places they chose are the same four that were generated by putting Yelp statistics through Silver's formula. Why not drop HRD to make room for El Castillito? Many of the reviews for HRD aren't even about the burrito.

                They did drop the doughnut shop with the bacon, egg, and hash browns burrito.



                1 Reply
                1. re: Robert Lauriston

                  The Doughnut Shop's contains bacon, egg and hash browns? By that definition, it would be the "true" Breakfast Burrito.

                2. And the western states are seeded:


                  and a short article on the cultural significance of the burrito:


                  5 Replies
                  1. re: kungful

                    I feel like this is rigged against SF. We have only Yelp-score picks in the running, including the obvious loser HRD. Southern California has a bunch of expert substitutions. The western states have five Yelp picks vs. 15 experts'.

                    1. re: Robert Lauriston

                      I think in order to get enough picks to make the brackets, they are having to put a large effort into the non-California regions.
                      They could have easily swamped the initial brackets with just the bay area, but had to make it look like a fair contest with the rest of the country. Perhaps they should have acted on their idea to make the four regions NorCal, SoCal, East and West (or North and South).

                      Simply by the nature of the contest (all regions on an even footed start), they are eliminating many of the better burritos. If there are a heavy concentration of quality burritos in one area, why not have more from that region ? (the answer is: because it makes for a better article if the brackets are pretty).

                      1. re: Robert Lauriston

                        They didn't come close to the best spots here in lalaland. But we're more a taco town anyways. But we do have this

                        1. re: Robert Lauriston

                          If I was going to be fair, I'd also say that, in my travels, burritos elsewhere in the country are pretty good these days. Mexicans established themselves everywhere that construction and gardeners were in need, so strong burritos in many municipalities.

                          This 538 thing is pretty poor though. Shows that Silver's mechanism of using "expert adjusted data" can fall down, with poor expert pruning.

                          1. re: bbulkow

                            It's just a fun statistical approach to something different if you ask me.

                      2. Okay, now that I've read several 538 pieces on the matter...

                        This entire idea, which even he admits is "crazy," is fundamentally flawed. Silver's primary skill is crunching data and projecting conclusions. It's basically straight math: I have this pool of data X, therefore conclusion Y. Both are flimsy in the burrito example.

                        First of all, the data pool is already suspect. There are a million reasons for this, but the short version is that his system is too reliant on the highly suspect assessments of random Yelp users. Having a professional burrito eater winnow the pool does not help: her subjective tastes will inherently color the results, making the final result less about "best burrito" and more about "the best burrito according to my buddies." Using a VORB formula doesn't help either; as Robert said, you are manipulating bad data to begin with.

                        Secondly, no such conclusion can readily be drawn from the data given. In both sports and politics, Silver is projecting a result among a narrow set of contestants in a specific contest -- e.g., either Obama or Romney will win this presidential election, or either the Rangers or Kings will win the NHL Finals. For both, the data utilized can be closely tied to the specific result he is projecting (e.g., "teams with longer puck possession tend to win games", "minorities tend to vote Democrat").

                        This cannot be said about the burrito makers. There is no statistical measure from which we can draw this conclusion -- e.g., we cannot say, "Oh, Taqueria Dondequiera is probably the best because their ratio of meat to bean is higher than everyone else's." The number of contestants only compounds the problem: projecting elections and games is a mutually exclusive scenario: e.g., voting for Obama inherently means you are not voting for Romney. This is obviously not the case with the burritos -- a person is allowed to like more than one burrito, and hasn't even tried 1% of all the burritos out there.

                        Honestly, I like Silver, but I suspect he is doing this for the same reason everyone else is doing it: this sort of thing generates hits. Probably he just thinks it's fun too, which is understandable, but surely he doesn't expect this to present an accurate result. What would have been more convincing is if he had first tried this method with Wicker Park, to see if the results were as he expected.

                        1 Reply
                        1. re: dunstable

                          Yes, the whole exercise is whimsical and crazy. But I think it is an interesting foil for examining the merits of his Bayesian approach. If you read through his handling of the Yelp data, he recognizes the flaws in the data and uses multiple independent measures to refine it. He also uses his domain knowledge to then drill into anomalies in his results to refine the method to improve the outcome. The Yelp analysis is pretty straight up Bayesian data mining. To me, its unfortunate that the statistical modeling only has Yelp data to work with, and ends with the bracket selections. It would be more interesting if he was able to layer in other data sets, or apply some level of metrics to the bracket scoring.

                          Ultimately, maybe it is just clickbait. There is certainly nothing definitive about it, by his own admission. But it is intrinsically more interesting than the standard best-of lists. Maybe we'll see him apply the same methodology to another dish - biscuits & gravy? croissants? fried chicken? Maybe he'll add new independent data sources to future analyses - Chow would do well to make their regional boards and DOTM threads available to him, for instance.