diff --git a/smtcomp/defs.py b/smtcomp/defs.py
index 05ecb3f2..d99c6204 100644
--- a/smtcomp/defs.py
+++ b/smtcomp/defs.py
@@ -1480,12 +1480,13 @@ def data(self) -> Path:
             raise ValueError("Configuration without data")
         return self._data
 
+    @functools.cached_property
+    def previous_years(self) -> list[int]:
+        return list(range(self.oldest_previous_results, self.current_year))
+
     @functools.cached_property
     def previous_results(self) -> list[tuple[int, Path]]:
-        return [
-            (year, self.data.joinpath(f"results-sq-{year}.json.gz"))
-            for year in range(Config.oldest_previous_results, Config.current_year)
-        ]
+        return [(year, self.data.joinpath(f"results-sq-{year}.json.gz")) for year in self.previous_years]
 
     @functools.cached_property
     def current_results(self) -> dict[Track, Path]:
diff --git a/smtcomp/results.py b/smtcomp/results.py
index c0a1d13a..c5696b96 100644
--- a/smtcomp/results.py
+++ b/smtcomp/results.py
@@ -446,6 +446,8 @@ def helper_get_results(
 
     -1 is used when no answer is available.
 
+    The second value returned is the selection
+
     """
     if len(results) == 0:
         lf = (
diff --git a/smtcomp/selection.py b/smtcomp/selection.py
index c1204dbe..3f77c354 100644
--- a/smtcomp/selection.py
+++ b/smtcomp/selection.py
@@ -164,7 +164,7 @@ def track_selection(benchmarks_with_info: pl.LazyFrame, config: defs.Config, tar
     )
 
     # Filter with the selected benchmarks
-    return benchmarks_with_info.join(selected_benchmarks, on="file", how="outer_coalesce").with_columns(
+    return benchmarks_with_info.join(selected_benchmarks, on="file", how="full").with_columns(
         pl.col("selected").fill_null(False)
     )
 
diff --git a/smtcomp/test_generation.py b/smtcomp/test_generation.py
new file mode 100644
index 00000000..42ab7fd3
--- /dev/null
+++ b/smtcomp/test_generation.py
@@ -0,0 +1,147 @@
+import smtcomp.defs as defs
+from pydantic.networks import HttpUrl
+from pathlib import Path
+from smtcomp.unpack import write_cin
+
+logic = defs.Logic.QF_ABV
+""" An arbitrary logic present in all tested tracks """
+
+solver_best = "SolverBest"
+""" Solver which prove everything """
+
+solver_sound = "SolverSound"
+""" Solver which is sound but don't solve everything """
+
+solver_error = "SolverError"
+""" Solver which makes errors """
+
+daisy_family = tuple(["daisy"])
+
+file_sat = defs.Smt2File(incremental=False, logic=defs.Logic.QF_ABV, family=daisy_family, name="sat.smt2")
+
+file_unsat = defs.Smt2File(incremental=False, logic=defs.Logic.QF_ABV, family=daisy_family, name="unsat.smt2")
+
+file_unknown = defs.Smt2File(incremental=False, logic=defs.Logic.QF_ABV, family=daisy_family, name="unkown.smt2")
+
+file_incremental = defs.Smt2File(
+    incremental=True, logic=defs.Logic.QF_ABV, family=daisy_family, name="incremental.smt2"
+)
+
+
+def mk_submissions() -> list[defs.Submission]:
+    contributor = "Author"
+    email = "Author@research.mars"
+
+    def mk_submission(name: str) -> defs.Submission:
+        return defs.Submission(
+            name=name,
+            contributors=[defs.Contributor(name=contributor)],
+            contacts=[defs.NameEmail(name=contributor, email=email)],
+            command=defs.Command(binary=name),
+            archive=defs.Archive(url=HttpUrl("http://research.mars/archive.tgz")),
+            website=HttpUrl("http://research.mars/"),
+            system_description=HttpUrl("http://research.mars/"),
+            solver_type=defs.SolverType.standalone,
+            participations=defs.Participations(
+                [
+                    defs.Participation(
+                        tracks=[defs.Track.SingleQuery, defs.Track.UnsatCore, defs.Track.Incremental],
+                        logics=defs.Logics([defs.Logic.QF_ABV]),
+                    )
+                ]
+            ),
+            seed=42,
+        )
+
+    return [
+        mk_submission(solver_best),
+        mk_submission(solver_sound),
+        mk_submission(solver_error),
+    ]
+
+
+def mk_benchmarks() -> defs.Benchmarks:
+    # sort file by name
+    return defs.Benchmarks(
+        non_incremental=[
+            defs.InfoNonIncremental(file=file_sat, status=defs.Status.Sat, asserts=10),
+            defs.InfoNonIncremental(file=file_unknown, status=defs.Status.Unknown, asserts=10),
+            defs.InfoNonIncremental(file=file_unsat, status=defs.Status.Unsat, asserts=10),
+        ],
+        incremental=[
+            defs.InfoIncremental(file=file_incremental, check_sats=10),
+        ],
+    )
+
+
+def mk_results(
+    track: defs.Track, solver: str, results: list[tuple[defs.Smt2File, defs.Answer, int]]
+) -> list[defs.Result]:
+    return [
+        defs.Result(
+            track=track,
+            solver=solver,
+            file=file,
+            result=result,
+            cpu_time=10.0,
+            wallclock_time=20.0,
+            memory_usage=10.0,
+            nb_answers=nb_answers,
+        )
+        for (file, result, nb_answers) in results
+    ]
+
+
+def mk_previous_years_results(config: defs.Config) -> list[tuple[int, Path, defs.Results]]:
+    results = [(file_sat, defs.Answer.Sat, 1)]
+    return [
+        (year, p, defs.Results(results=mk_results(defs.Track.SingleQuery, solver_best, results)))
+        for (year, p) in config.previous_results
+    ]
+
+
+def mk_sq_results() -> defs.Results:
+    results = [(file_sat, defs.Answer.Sat, 1), (file_unknown, defs.Answer.Unsat, 1), (file_unsat, defs.Answer.Unsat, 1)]
+
+    return defs.Results(
+        results=mk_results(defs.Track.SingleQuery, solver_best, results)
+        + mk_results(defs.Track.SingleQuery, solver_sound, results[:-1] + [(file_unsat, defs.Answer.Unknown, 1)])
+        + mk_results(defs.Track.SingleQuery, solver_error, results[:-1] + [(file_unsat, defs.Answer.Sat, 1)])
+    )
+
+
+def mk_uc_results() -> defs.Results:
+    results = [(file_unknown, defs.Answer.Unsat, 7), (file_unsat, defs.Answer.Unsat, 7)]
+
+    return defs.Results(
+        results=mk_results(defs.Track.SingleQuery, solver_best, results)
+        + mk_results(defs.Track.SingleQuery, solver_sound, results[:-1] + [(file_unsat, defs.Answer.Unknown, 1)])
+        + mk_results(defs.Track.SingleQuery, solver_error, results[:-1] + [(file_unsat, defs.Answer.Sat, 1)])
+    )
+
+
+def mk_inc_results() -> defs.Results:
+    results = [(file_incremental, defs.Answer.Incremental, 7)]
+
+    return defs.Results(
+        results=mk_results(defs.Track.SingleQuery, solver_best, results)
+        + mk_results(defs.Track.SingleQuery, solver_sound, [(file_incremental, defs.Answer.Incremental, 5)])
+        + mk_results(defs.Track.SingleQuery, solver_error, [(file_incremental, defs.Answer.IncrementalError, 0)])
+    )
+
+
+def write_test_files(data: Path) -> None:
+    config = defs.Config(data)
+    submissions = data / ".." / "submissions"
+    submissions.mkdir()
+    for submission in mk_submissions():
+        (submissions / (submission.name + ".json")).write_text(submission.model_dump_json(indent=2))
+    write_cin(config.benchmarks, mk_benchmarks().model_dump_json())
+    for _, file, results in mk_previous_years_results(config):
+        write_cin(file, results.model_dump_json(indent=2))
+    for track, gen in [
+        (defs.Track.SingleQuery, mk_sq_results),
+        (defs.Track.UnsatCore, mk_uc_results),
+        (defs.Track.Incremental, mk_inc_results),
+    ]:
+        write_cin(config.current_results[track], gen().model_dump_json(indent=2))
diff --git a/tests/test_validate.py b/tests/test_validate.py
index 5fca6113..dfbcf7bb 100644
--- a/tests/test_validate.py
+++ b/tests/test_validate.py
@@ -10,6 +10,9 @@
 from smtcomp.submission import read
 from smtcomp.generate_benchmarks import generate_trivial_benchmarks, path_trivial_benchmark
 
+import smtcomp.test_generation as gtests
+import smtcomp.results, smtcomp.scoring, smtcomp.generate_website_page
+
 runner = CliRunner()
 good_cases = ["tests/test1.json", "submissions/template/template.json", "submissions/template/template_aws_tracks.json"]
 bad_cases = ["test_bad.json"]
@@ -56,3 +59,89 @@ def test_generate_trivial(tmp_path: Path) -> None:
             for logic in logics:
                 for status in statuses:
                     assert path_trivial_benchmark(tmp_path, track, logic, status).exists()
+
+
+@pytest.fixture(scope="session")
+def config(tmp_path_factory: pytest.TempPathFactory) -> defs.Config:
+    fn = tmp_path_factory.mktemp("tmp")
+    data = fn / "data"
+    data.mkdir()
+    gtests.write_test_files(data)
+    result = runner.invoke(app, ["create-cache", str(data.absolute())])
+    assert result.exit_code == 0
+    return defs.Config(data)
+
+
+def test_results_sq_export(config: defs.Config) -> None:
+    print(config.data)
+    track = defs.Track.SingleQuery
+    results, selection = smtcomp.results.helper_get_results(config, [], track)
+    scores = smtcomp.scoring.add_disagreements_info(results, track)
+    sound_solvers = scores.filter(sound_solver=True).select("solver").unique().collect()["solver"].sort().to_list()
+    assert sound_solvers == [gtests.solver_best, gtests.solver_sound]
+    smtcomp.generate_website_page.export_results(config, selection, results, track)
+    page_suffix = smtcomp.generate_website_page.page_track_suffix(track)
+    podium = smtcomp.generate_website_page.PodiumDivision.model_validate_json(
+        (config.web_results / f"{gtests.logic.name.lower()}-{page_suffix}.md").read_text()
+    )
+    print(podium)
+    assert podium.winner_seq == gtests.solver_best
+
+    assert podium.sequential[0].name == gtests.solver_best
+    assert podium.sequential[0].errorScore == 0
+    assert podium.sequential[0].correctScore == 3
+    assert podium.sequential[1].name == gtests.solver_sound
+    assert podium.sequential[1].errorScore == 0
+    assert podium.sequential[1].correctScore == 2
+    assert podium.sequential[2].name == gtests.solver_error
+    assert podium.sequential[2].errorScore == 1
+    assert podium.sequential[2].correctScore == 2
+
+
+def test_results_uc_export(config: defs.Config) -> None:
+    print(config.data)
+    track = defs.Track.UnsatCore
+    results, selection = smtcomp.results.helper_get_results(config, [], track)
+    scores = smtcomp.scoring.add_disagreements_info(results, track)
+    sound_solvers = scores.filter(sound_solver=True).select("solver").unique().collect()["solver"].sort().to_list()
+    assert sound_solvers == [gtests.solver_best, gtests.solver_sound]
+    smtcomp.generate_website_page.export_results(config, selection, results, track)
+    page_suffix = smtcomp.generate_website_page.page_track_suffix(track)
+    podium = smtcomp.generate_website_page.PodiumDivision.model_validate_json(
+        (config.web_results / f"{gtests.logic.name.lower()}-{page_suffix}.md").read_text()
+    )
+    print(podium)
+    assert podium.winner_seq == gtests.solver_best
+
+    assert podium.sequential[0].name == gtests.solver_best
+    assert podium.sequential[0].errorScore == 0
+    assert podium.sequential[0].correctScore == 6
+    assert podium.sequential[1].name == gtests.solver_sound
+    assert podium.sequential[1].errorScore == 0
+    assert podium.sequential[1].correctScore == 3
+    assert podium.sequential[2].name == gtests.solver_error
+    assert podium.sequential[2].errorScore == 1
+    assert podium.sequential[2].correctScore == 3
+
+
+def test_results_inc_export(config: defs.Config) -> None:
+    print(config.data)
+    track = defs.Track.Incremental
+    results, selection = smtcomp.results.helper_get_results(config, [], track)
+    smtcomp.generate_website_page.export_results(config, selection, results, track)
+    page_suffix = smtcomp.generate_website_page.page_track_suffix(track)
+    podium = smtcomp.generate_website_page.PodiumDivision.model_validate_json(
+        (config.web_results / f"{gtests.logic.name.lower()}-{page_suffix}.md").read_text()
+    )
+    print(podium)
+    assert podium.winner_seq == gtests.solver_best
+
+    assert podium.sequential[0].name == gtests.solver_best
+    assert podium.sequential[0].errorScore == 0
+    assert podium.sequential[0].correctScore == 7
+    assert podium.sequential[1].name == gtests.solver_sound
+    assert podium.sequential[1].errorScore == 0
+    assert podium.sequential[1].correctScore == 5
+    assert podium.sequential[2].name == gtests.solver_error
+    assert podium.sequential[2].errorScore == 1
+    assert podium.sequential[2].correctScore == 0